Endo, Kaori (Kawasaki, JP)
Matsumoto, Chikako (Kawasaki, JP)
Togawa, Taro (Kawasaki, JP)
Ota, Yasuji (Kawasaki, JP)

11/802228

04/17/2008

05/21/2007

Images are available in PDF form when logged in. To view PDFs, Login  or  Create Account (Free!)

View patents that cite this patent

Click for automatic bibliography generation

Fujitsu Limited (Kawasaki, JP)

704/207

704/E21.002

G10L11/04

STAAS & HALSEY LLP (SUITE 700, 1201 NEW YORK AVENUE, N.W., WASHINGTON, DC, 20005, US)

What is claimed is:

1. A pitch conversion method comprising: a degradation evaluation step of inputting an input signal pitch pattern per predetermined processing unit and a target pitch pattern for the input signal pitch pattern, and of calculating a degradation degree indicating how a waveform of the input signal degrades upon pitch conversion from the input signal pitch pattern to the target pitch pattern; and a pitch conversion step of performing the pitch conversion with predetermined data throughput depending on the degradation degree.

2. The pitch conversion method as claimed in claim 1, wherein the degradation evaluation step includes an average pitch conversion amount calculation step of calculating an average pitch conversion amount by dividing a sum of pitch differences between the target pitch pattern and the input signal pitch pattern per predetermined cycle by a sum of pitches of the input signal pitch pattern per predetermined cycle, and a degradation degree calculation step of providing as the degradation degree to the pitch conversion step a value that is the average pitch conversion amount weighted by predetermined coefficients.

3. The pitch conversion method as claimed in claim 1, wherein the pitch conversion step includes a first and second pitch conversion steps depending on a level of the degradation degree, accordingly the degradation evaluation step also includes the identical first and second pitch conversion steps, and the degradation evaluation step further includes an average signal difference calculation step of calculating an average signal difference by dividing a sum of power differences between a first pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the first pitch conversion step and a second pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the second pitch conversion step per predetermined cycle by a sum of powers of the second pitch conversion result per predetermined cycle, and a degradation degree calculation step of providing as the degradation degree to the pitch conversion step a value that is the average signal difference weighted by predetermined coefficients.

4. The pitch conversion method as claimed in claim 1, wherein the degradation evaluation step includes a pitch pattern change degree calculation step of classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and of determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and a degradation degree calculation step of providing as the degradation degree to the pitch conversion step a value that is the pitch pattern change degree weighted by predetermined coefficients.

5. The pitch conversion method as claimed in claim 2, wherein the pitch conversion step includes a first and second pitch conversion steps depending on a level of the degradation degree, accordingly the degradation evaluation step also includes the identical first and second pitch conversion steps, and the degradation evaluation step further includes an average signal difference calculation step of calculating an average signal difference by dividing a sum of power differences between a first pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the first pitch conversion step and a second pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the second pitch conversion step per predetermined cycle by a sum of powers of the second pitch conversion result per predetermined cycle, and the degradation degree calculation step provides as the degradation degree to the pitch conversion step a sum of values that are the average pitch conversion amount and the average signal difference respectively weighted by predetermined coefficients.

6. The pitch conversion method as claimed in claim 2, wherein the degradation evaluation step further includes a pitch pattern change degree calculation step of classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and of determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and the degradation degree calculation step provides as the degradation degree to the pitch conversion step a sum of values that are the average pitch conversion amount and the pitch pattern change degree respectively weighted by predetermined coefficients.

7. The pitch conversion method as claimed in claim 3, wherein the degradation evaluation step further includes a pitch pattern change degree calculation step of classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and of determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and the degradation degree calculation step provides as the degradation degree to the pitch conversion step a sum of values that are the average signal difference and the pitch pattern change degree respectively weighted by predetermined coefficients.

8. The pitch conversion method as claimed in claim 5, wherein the degradation evaluation step further includes a pitch pattern change degree calculation step of classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and of determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and the degradation degree calculation step provides as the degradation degree to the pitch conversion step a sum of values that are the average pitch conversion amount, the average signal difference, and the pitch pattern change degree respectively weighted by predetermined coefficients.

9. A pitch conversion method comprising: a degradation degree extraction step of inputting a voice state and a phonemic type of an input signal per predetermined processing unit, and of extracting a degradation degree corresponding to the voice state and the phonemic type inputted from a database in which degradation degrees indicating how a waveform of the input signal degrades upon pitch conversion from an input signal pitch pattern to a target pitch pattern for the input signal pitch pattern are associated with all of combinations of voice states and phonemic types estimated to be recorded; and a pitch conversion step of performing the pitch conversion with predetermined data throughput depending on the degradation degree.

10. A pitch conversion method comprising: a degradation evaluation step of inputting an input signal pitch pattern per predetermined processing unit, a target pitch pattern for the input signal pitch pattern, and a voice state and a phonemic type of the input signal, and of calculating a degradation degree indicating how a waveform of the input signal degrades upon pitch conversion from the input signal pitch pattern to the target pitch pattern; and a pitch conversion step of performing the pitch conversion with predetermined data throughput depending on the degradation degree.

11. The pitch conversion method as claimed in claim 10, wherein the degradation evaluation step includes an average pitch conversion amount calculation step of calculating an average pitch conversion amount by dividing a sum of pitch differences between the target pitch pattern and the input signal pitch pattern per predetermined cycle by a sum of pitches of the input signal pitch pattern per predetermined cycle, a degradation degree extraction step of extracting a degradation degree corresponding to the voice state and the phonemic type inputted from a database in which the degradation degrees are associated with all of combinations of voice states and phonemic types estimated to be recorded, and a degradation degree calculation step of providing as the degradation degree to the pitch conversion step a sum of values that are the average pitch conversion amount and the extracted degradation degree respectively weighted by predetermined coefficients.

12. The pitch conversion method as claimed in claim 10, wherein the pitch conversion step includes a first and second pitch conversion steps depending on a level of the degradation degree, accordingly the degradation evaluation step also includes the identical first and second pitch conversion steps, and the degradation evaluation step further includes an average signal difference calculation step of calculating an average signal difference by dividing a sum of power differences between a first pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the first pitch conversion step and a second pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the second pitch conversion step per predetermined cycle by a sum of powers of the second pitch conversion result per predetermined cycle, a degradation degree extraction step of extracting a degradation degree corresponding to the voice state and the phonemic type inputted from a database in which the degradation degrees are associated with all of combinations of voice states and phonemic types estimated to be recorded, and a degradation degree calculation step of providing as the degradation degree to the pitch conversion step a sum of values that are the average signal difference and the extracted degradation degree respectively weighted by predetermined coefficients.

13. The pitch conversion method as claimed in claim 10, wherein the degradation evaluation step includes a pitch pattern change degree calculation step of classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and of determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, a degradation degree extraction step of extracting a degradation degree corresponding to the voice state and the phonemic type inputted from a database in which the degradation degrees are associated with all of combinations of voice states and phonemic types estimated to be recorded, and a degradation degree calculation step of providing as the degradation degree to the pitch conversion step a sum of values that are the pitch pattern change degree and the extracted degradation degree respectively weighted by predetermined coefficients.

14. The pitch conversion method as claimed in claim 11, wherein the pitch conversion step includes a first and second pitch conversion steps depending on a level of the degradation degree, accordingly the degradation evaluation step also includes the identical first and second pitch conversion steps, and the degradation evaluation step further includes an average signal difference calculation step of calculating an average signal difference by dividing a sum of power differences between a first pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the first pitch conversion step and a second pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the second pitch conversion step per predetermined cycle by a sum of powers of the second pitch conversion result per predetermined cycle, and the degradation degree calculation step provides as the degradation degree to the pitch conversion step a sum of values that are the average pitch conversion amount, the extracted degradation degree, and the average signal difference respectively weighted by predetermined coefficients.

15. The pitch conversion method as claimed in claim 11, wherein the degradation evaluation step further includes a pitch pattern change degree calculation step of classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and of determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and a degradation degree calculation step provides as the degradation degree to the pitch conversion step a sum of values that are the average pitch conversion amount, the extracted degradation degree, and the pitch pattern change degree respectively weighted by predetermined coefficients.

16. The pitch conversion method as claimed in claim 12, wherein the degradation evaluation step further includes a pitch pattern change degree calculation step of classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and of determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and a degradation degree calculation step provides as the degradation degree to the pitch conversion step a sum of values that are the average signal difference, the extracted degradation degree, and the pitch pattern change degree respectively weighted by predetermined coefficients.

17. The pitch conversion method as claimed in claim 14, wherein the degradation evaluation step further includes a pitch pattern change degree calculation step of classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and of determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and a degradation degree calculation step provides as the degradation degree to the pitch conversion step a sum of values that are the average pitch conversion amount, the extracted degradation degree, the average signal difference, and the pitch pattern change degree respectively weighted by predetermined coefficients.

18. A pitch conversion device comprising: a degradation evaluation means inputting an input signal pitch pattern per predetermined processing unit and a target pitch pattern for the input signal pitch pattern, and calculating a degradation degree indicating how a waveform of the input signal degrades upon pitch conversion from the input signal pitch pattern to the target pitch pattern; and a pitch conversion means performing the pitch conversion with predetermined data throughput depending on the degradation degree.

19. The pitch conversion device as claimed in claim 18, wherein the degradation evaluation means includes an average pitch conversion amount calculation means calculating an average pitch conversion amount by dividing a sum of pitch differences between the target pitch pattern and the input signal pitch pattern per predetermined cycle by a sum of pitches of the input signal pitch pattern per predetermined cycle, and a degradation degree calculation means providing as the degradation degree to the pitch conversion means a value that is the average pitch conversion amount weighted by predetermined coefficients.

20. The pitch conversion device as claimed in claim 18, wherein the pitch conversion means includes a first and second pitch conversion means depending on a level of the degradation degree, accordingly the degradation evaluation means also includes the identical first and second pitch conversion means, and the degradation evaluation means further includes an average signal difference calculation means calculating an average signal difference by dividing a sum of power differences between a first pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the first pitch conversion means and a second pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the second pitch conversion means per predetermined cycle by a sum of powers of the second pitch conversion result per predetermined cycle, and a degradation degree calculation means providing as the degradation degree to the pitch conversion means a value that is the average signal difference weighted by predetermined coefficients.

21. The pitch conversion device as claimed in claim 18, wherein the degradation evaluation means includes a pitch pattern change degree calculation means classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and a degradation degree calculation means providing as the degradation degree to the pitch conversion means a value that is the pitch pattern change degree weighted by predetermined coefficients.

22. The pitch conversion device as claimed in claim 19, wherein the pitch conversion means includes a first and second pitch conversion means depending on a level of the degradation degree, accordingly the degradation evaluation means also includes the identical first and second pitch conversion means, and the degradation evaluation means further includes an average signal difference calculation means calculating an average signal difference by dividing a sum of power differences between a first pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the first pitch conversion means and a second pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the second pitch conversion means per predetermined cycle by a sum of powers of the second pitch conversion result per predetermined cycle, and the degradation degree calculation means provides as the degradation degree to the pitch conversion means a sum of values that are the average pitch conversion amount and the average signal difference respectively weighted by predetermined coefficients.

23. The pitch conversion device as claimed in claim 19, wherein the degradation evaluation means further includes a pitch pattern change degree calculation means classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and the degradation degree calculation means provides as the degradation degree to the pitch conversion means a sum of values that are the average pitch conversion amount and the pitch pattern change degree respectively weighted by predetermined coefficients.

24. The pitch conversion device as claimed in claim 20, wherein the degradation evaluation means further includes a pitch pattern change degree calculation means classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and the degradation degree calculation means provides as the degradation degree to the pitch conversion means a sum of values that are the average signal difference and the pitch pattern change degree respectively weighted by predetermined coefficients.

25. The pitch conversion device as claimed in claim 22, wherein the degradation evaluation means further includes a pitch pattern change degree calculation means classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and the degradation degree calculation means provides as the degradation degree to the pitch conversion means a sum of values that are the average pitch conversion amount, the average signal difference, and the pitch pattern change degree respectively weighted by predetermined coefficients.

26. A pitch conversion device comprising: a degradation degree extraction means inputting a voice state and a phonemic type of an input signal per predetermined processing unit, and extracting a degradation degree corresponding to the voice state and the phonemic type inputted from a database in which degradation degrees indicating how a waveform of the input signal degrades upon pitch conversion from an input signal pitch pattern to a target pitch pattern for the input signal pitch pattern are associated with all of combinations of voice states and phonemic types estimated to be recorded; and a pitch conversion means performing the pitch conversion with predetermined data throughput depending on the degradation degree.

27. A pitch conversion device comprising: a degradation evaluation means inputting an input signal pitch pattern per predetermined processing unit, a target pitch pattern for the input signal pitch pattern, and a voice state and a phonemic type of the input signal, and calculating a degradation degree indicating how a waveform of the input signal degrades upon pitch conversion from the input signal pitch pattern to the target pitch pattern; and a pitch conversion means performing the pitch conversion with predetermined data throughput depending on the degradation degree.

28. The pitch conversion device as claimed in claim 27, wherein the degradation evaluation means includes an average pitch conversion amount calculation means calculating an average pitch conversion amount by dividing a sum of pitch differences between the target pitch pattern and the input signal pitch pattern per predetermined cycle by a sum of pitches of the input signal pitch pattern per predetermined cycle, a degradation degree extraction means extracting a degradation degree corresponding to the voice state and the phonemic type inputted from a database in which the degradation degrees are associated with all of combinations of voice states and phonemic types estimated to be recorded, and a degradation degree calculation means providing as the degradation degree to the pitch conversion means a sum of values that are the average pitch conversion amount and the extracted degradation degree respectively weighted by predetermined coefficients.

29. The pitch conversion device as claimed in claim 27, wherein the pitch conversion means includes a first and second pitch conversion means depending on a level of the degradation degree, accordingly the degradation evaluation means also includes the identical first and second pitch conversion means, and the degradation evaluation means further includes an average signal difference calculation means calculating an average signal difference by dividing a sum of power differences between a first pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the first pitch conversion means and a second pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the second pitch conversion means per predetermined cycle by a sum of powers of the second pitch conversion result per predetermined cycle, a degradation degree extraction means extracting a degradation degree corresponding to the voice state and the phonemic type inputted from a database in which the degradation degrees are associated with all of combinations of voice states and phonemic types estimated to be recorded, and a degradation degree calculation means providing as the degradation degree to the pitch conversion means a sum of values that are the average signal difference and the extracted degradation degree respectively weighted by predetermined coefficients.

30. The pitch conversion device as claimed in claim 27, wherein the degradation evaluation means includes a pitch pattern change degree calculation means classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, a degradation degree extraction means extracting a degradation degree corresponding to the voice state and the phonemic type inputted from a database in which the degradation degrees are associated with all of combinations of voice states and phonemic types estimated to be recorded, and a degradation degree calculation means providing as the degradation degree to the pitch conversion means a sum of values that are the pitch pattern change degree and the extracted degradation degree respectively weighted by predetermined coefficients.

31. The pitch conversion device as claimed in claim 28, wherein the pitch conversion means includes a first and second pitch conversion means depending on a level of the degradation degree, accordingly the degradation evaluation means also includes the identical first and second pitch conversion means, and the degradation evaluation means further includes an average signal difference calculation means calculating an average signal difference by dividing a sum of power differences between a first pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the first pitch conversion means and a second pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the second pitch conversion means per predetermined cycle by a sum of powers of the second pitch conversion result per predetermined cycle, and the degradation degree calculation means provides as the degradation degree to the pitch conversion means a sum of values that are the average pitch conversion amount, the extracted degradation degree, and the average signal difference respectively weighted by predetermined coefficients.

32. The pitch conversion device as claimed in claim 28, wherein the degradation evaluation means further includes a pitch pattern change degree calculation means classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and a degradation degree calculation means provides as the degradation degree to the pitch conversion means a sum of values that are the average pitch conversion amount, the extracted degradation degree, and the pitch pattern change degree respectively weighted by predetermined coefficients.

33. The pitch conversion device as claimed in claim 29, wherein the degradation evaluation means further includes a pitch pattern change degree calculation means classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and a degradation degree calculation means provides as the degradation degree to the pitch conversion means a sum of values that are the average signal difference, the extracted degradation degree, and the pitch pattern change degree respectively weighted by predetermined coefficients.

34. The pitch conversion device as claimed in claim 31, wherein the degradation evaluation means further includes a pitch pattern change degree calculation means classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and a degradation degree calculation means provides as the degradation degree to the pitch conversion means a sum of values that are the average pitch conversion amount, the extracted degradation degree, the average signal difference, and the pitch pattern change degree respectively weighted by predetermined coefficients.

BACKGROUND OF THE INVENTION

1. Field of the Invention

The present invention relates to a pitch conversion method and device, and in particular to a pitch conversion method and device for converting a pitch of an input signal into a desired (target) pitch in order to change e.g. a voice level or accent.

2. Description of the Related Art

Prior art examples [1] and [2] of the above-mentioned pitch conversion technology will now be described referring to FIG. 24.

Prior Art Example [1] (PSOLA method): FIG. 24

In this pitch conversion technology, a pitch conversion is performed by overlapping and adding waveforms of an input signal per pitch cycle in conformity with a target pitch (namely, the input signal is eventually expanded or contracted in the direction of time axis), and is generally called a PSOLA (Pitch-Synchronous Overlap and Add) method (see e.g. patent document 1).

FIG. 24 shows an example of the pitch conversion for contracting an input signal “In” in the direction of time axis by using the PSOLA method.

Namely, two waveforms W 1 and W 2 are firstly cut from the input signal “In” per pitch cycle T, and then window functions F 1 and F 2 are respectively applied to the cut waveforms W 1 and W 2 to adjust the amplitudes. In order to avoid discontinuity of waveforms at the boundary between an overlapped portion of the waveforms W 1 and W 2 by overlapping and adding which will be described later and the non-overlapped portion, the window functions F 1 and F 2 are set so that the sum of mutual contribution degrees may become “1” at the overlapped portion of the waveforms W 1 and W 2 as shown in FIG. 24.

Then, two waveforms (not shown) whose amplitudes are adjusted by the window functions F 1 and F 2 are overlapped and added to obtain the output signal “Out”.

In such a prior art example [1], waveforms after the pitch conversion may be deformed since waveforms whose phases are different from each other are overlapped. This deformation is notable especially when a pitch conversion ratio (namely, an expansion and contraction ratio of the input signal in the direction of time axis) is large, which leads to a degradation of sound quality.

In order to deal with this problem, a prior art example [2] has been already proposed as described herebelow:

Prior Art Example [2]: Not shown

In this pitch conversion technology, a linear predictive analysis is firstly performed to the input signal, so that the signal is separated into an envelope signal (formant component) and a residual signal (harmonics component). Then, a pitch conversion is performed only to the residual signal in the same way as the above-mentioned prior art example [1], so that the residual signal after the pitch conversion has been performed and the original envelope signal are synthesized by using a linear predictive coefficient calculated from the input signal.

Thus, the pitch conversion can be performed without affecting the envelope signal, and the above-mentioned waveform deformation due to the pitch conversion can be reduced, so that a degradation of sound quality can be avoided (see e.g. patent document 2).

[Patent document 1] Japanese Patent Application Laid-open No. 10-78791

[Patent document 2] Japanese Patent Application Laid-open No. 7-219597

While in the above-mentioned prior art example [2] the pitch conversion can be performed without deteriorating the sound quality of the input signal compared with the above-mentioned prior art example [1], there is a problem that the linear predictive analysis and the signal separation/synthesis require processing of large data throughput (calculation amount or the like).

SUMMARY OF THE INVENTION

It is accordingly an object of the present invention to provide a pitch conversion method and device which can reduce data throughput while suppressing a degradation of sound quality due to a pitch conversion as much as possible.

[1] In order to achieve the above-mentioned object, a pitch conversion method (or device) according to one aspect of the present invention comprises: a degradation evaluation step of (or means) inputting an input signal pitch pattern per predetermined processing unit and a target pitch pattern for the input signal pitch pattern, and of calculating a degradation degree indicating how a waveform of the input signal degrades upon pitch conversion from the input signal pitch pattern to the target pitch pattern; and a pitch conversion step of (or means) performing the pitch conversion with predetermined data throughput depending on the degradation degree.

Namely, at a degradation evaluation step (or means), a degradation degree is calculated in advance of the execution of a pitch conversion, and at a pitch conversion step (or means), data throughput for performing the pitch conversion is switched over depending on the degradation degree.

Thus, when the degradation degree is small, the pitch conversion can be performed with small data throughput by using the pitch conversion technology shown in e.g. the above-mentioned prior art example [1] since a degradation of sound quality due to the pitch conversion does not occur. Also, only when a high-performance pitch conversion is required to be performed due to a large degradation degree, the pitch conversion can be performed by using the pitch conversion technology shown in e.g. the above-mentioned prior art example [2]. Therefore, it is possible to reduce a processing load (i.e. the entire data throughput).

[2] Also, in the above-mentioned [1], the degradation evaluation step (or means) may include an average pitch conversion amount calculation step of (or means) calculating an average pitch conversion amount by dividing a sum of pitch differences between the target pitch pattern and the input signal pitch pattern per predetermined cycle by a sum of pitches of the input signal pitch pattern per predetermined cycle, and a degradation degree calculation step of (or means) providing as the degradation degree to the pitch conversion step (or means) a value that is the average pitch conversion amount weighted by predetermined coefficients.

Namely, since this average pitch conversion amount is a value indicating how much pitch conversion is required to be performed for an input signal per predetermined processing unit (namely, how a waveform of an input signal can be deformed), the value can be used as the degradation degree.

[3] Also, in the above-mentioned [1], the pitch conversion step (or means) may include a first and second pitch conversion steps (or means) depending on a level of the degradation degree, accordingly the degradation evaluation step (or means) may also include the identical first and second pitch conversion steps (or means), and the degradation evaluation step (or means) may further include an average signal difference calculation step of (or means) calculating an average signal difference by dividing a sum of power differences between a first pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the first pitch conversion step (or means) and a second pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the second pitch conversion step (or means) per predetermined cycle by a sum of powers of the second pitch conversion result per predetermined cycle, and a degradation degree calculation step of (or means) providing as the degradation degree to the pitch conversion step (or means) a value that is the average signal difference weighted by predetermined coefficients.

Namely, the degradation evaluation step (or means) performs the pitch conversion to the part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern in advance of the execution of the pitch conversion at a subsequent pitch conversion step (or means) respectively at the first pitch conversion step (or means) and the second pitch conversion step (or means) which are the same as the pitch conversion step (or means) included at the subsequent stage.

An average signal difference obtained based on the results of both pitch conversions mentioned above is a value indicating a difference closer to a difference between the results of the pitch conversions as respectively and actually performed at the first pitch conversion step (or means) and the second pitch conversion step (or means) included in the pitch conversion step (or means). When the average signal difference is small, it can be regarded that there is no difference between the pitch conversion results regardless of the size of data throughput (namely, the degradation of sound quality due to the pitch conversion does not occur regardless of the size of the data throughput). Therefore, the average signal difference can be used as the degradation degree.

[4] Also, in the above-mentioned [1], the degradation evaluation step (or means) may include a pitch pattern change degree calculation step of (or means) classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and of determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and a degradation degree calculation step of (or means) providing as the degradation degree to the pitch conversion step (or means) a value that is the pitch pattern change degree weighted by predetermined coefficients.

Namely, since this pitch pattern change degree is a value obtained from a correlation between the change trend of the input signal pitch pattern and that of the target pitch pattern (namely, e.g. a value indicating whether or not the pitch of the input signal is required to be greatly changed), the value can be used as the degradation degree.

[5] Also, in the above-mentioned [2], the pitch conversion step (or means) may include a first and second pitch conversion steps (or means) depending on a level of the degradation degree, accordingly the degradation evaluation step (or means) may also include the identical first and second pitch conversion steps (or means), and the degradation evaluation step (or means) may further include an average signal difference calculation step of (or means) calculating an average signal difference by dividing a sum of power differences between a first pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the first pitch conversion step (or means) and a second pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the second pitch conversion step (or means) per predetermined cycle by a sum of powers of the second pitch conversion result per predetermined cycle, and the degradation degree calculation step (or means) may provide as the degradation degree to the pitch conversion step (or means) a sum of values that are the average pitch conversion amount and the average signal difference respectively weighted by predetermined coefficients.

[6] Also, in the above-mentioned [2], the degradation evaluation step (or means) may further include a pitch pattern change degree calculation step of (or means) classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and of determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and the degradation degree calculation step (or means) may provide as the degradation degree to the pitch conversion step (or means) a sum of values that are the average pitch conversion amount and the pitch pattern change degree respectively weighted by predetermined coefficients.

[7] Also, in the above-mentioned [3], the degradation evaluation step (or means) may further include a pitch pattern change degree calculation step of (or means) classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and of determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and the degradation degree calculation step (or means) may provide as the degradation degree to the pitch conversion step (or means) a sum of values that are the average signal difference and the pitch pattern change degree respectively weighted by predetermined coefficients.

[8] Also, in the above-mentioned [5], the degradation evaluation step (or means) may further include a pitch pattern change degree calculation step of (or means) classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and of determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and the degradation degree calculation step (or means) may provide as the degradation degree to the pitch conversion step (or means) a sum of values that are the average pitch conversion amount, the average signal difference, and the pitch pattern change degree respectively weighted by predetermined coefficients.

As the above-mentioned [5]-[8], the combination of two or three of the average pitch conversion amount, the average signal difference, and the pitch pattern change degree described in the above-mentioned [2]-[4] can be used as the degradation degree.

[9] Also, a pitch conversion method (or device) according to one aspect of the present invention comprises: a degradation degree extraction step of (or means) inputting a voice state and a phonemic type of an input signal per predetermined processing unit, and extracting a degradation degree corresponding to the voice state and the phonemic type inputted from a database in which degradation degrees indicating how a waveform of the input signal degrades upon pitch conversion from an input signal pitch pattern to a target pitch pattern for the input signal pitch pattern are associated with all of combinations of voice states and phonemic types estimated to be recorded; and a pitch conversion step of (or means) performing the pitch conversion with predetermined data throughput depending on the degradation degree.

Namely, in this database, the combination of all of the voice states and the phonemic types estimated as the input signal are associated with the degradation degree to be recorded. Therefore, it is possible to accurately reduce the data throughput depending on the degradation of the sound quality which may actually occur.

[10] Also, a pitch conversion method (or device) according to one aspect of the present invention comprises: a degradation evaluation step of (or means) inputting an input signal pitch pattern per predetermined processing unit, a target pitch pattern for the input signal pitch pattern, and a voice state and a phonemic type of the input signal, and calculating a degradation degree indicating how a waveform of the input signal degrades upon pitch conversion from the input signal pitch pattern to the target pitch pattern; and a pitch conversion step of (or means) performing the pitch conversion with predetermined data throughput depending on the degradation degree.

Thus, the degradation degree can be calculated in consideration of both of the degradation degree based on the input signal pitch pattern and the target pitch pattern as described in the above-mentioned [1], and the degradation degree based on the voice state and the phonemic type of the input signal as described in the above-mentioned [9], thereby enabling the data throughput for the pitch conversion to be more accurately reduced while the degradation of sound quality is suppressed.

[11] Also, in the above-mentioned [10], the degradation evaluation step (or means) may include an average pitch conversion amount calculation step of (or means) calculating an average pitch conversion amount by dividing a sum of pitch differences between the target pitch pattern and the input signal pitch pattern per predetermined cycle by a sum of pitches of the input signal pitch pattern per predetermined cycle, a degradation degree extraction step of (or means) extracting a degradation degree corresponding to the voice state and the phonemic type inputted from a database in which the degradation degrees are associated with all of combinations of voice states and phonemic types estimated to be recorded, and a degradation degree calculation step of (or means) providing as the degradation degree to the pitch conversion step (or means) a sum of values that are the average pitch conversion amount and the extracted degradation degree respectively weighted by predetermined coefficients.

[12] Also, in the above-mentioned [10], the pitch conversion step (or means) may include a first and second pitch conversion steps (or means) depending on a level of the degradation degree, accordingly the degradation evaluation step (or means) may also include the identical first and second pitch conversion steps (or means), and the degradation evaluation step (or means) may further include an average signal difference calculation step of (or means) calculating an average signal difference by dividing a sum of power differences between a first pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the first pitch conversion step (or means) and a second pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the second pitch conversion step (or means) per predetermined cycle by a sum of powers of the second pitch conversion result per predetermined cycle, a degradation degree extraction step of (or means) extracting a degradation degree corresponding to the voice state and the phonemic type inputted from a database in which the degradation degrees are associated with all of combinations of voice states and phonemic types estimated to be recorded, and a degradation degree calculation step of (or means) providing as the degradation degree to the pitch conversion step (or means) a sum of values that are the average signal difference and the extracted degradation degree respectively weighted by predetermined coefficients.

[13] Also, in the above-mentioned [10], the degradation evaluation step (or means) may include a pitch pattern change degree calculation step of (or means) classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, a degradation degree extraction step of (or means) extracting a degradation degree corresponding to the voice state and the phonemic type inputted from a database in which the degradation degrees are associated with all of combinations of voice states and phonemic types estimated to be recorded, and a degradation degree calculation step of (or means) providing as the degradation degree to the pitch conversion step (or means) a sum of values that are the pitch pattern change degree and the extracted degradation degree respectively weighted by predetermined coefficients.

[14] Also, in the above-mentioned [11], the pitch conversion step (or means) may include a first and second pitch conversion steps (or means) depending on a level of the degradation degree, accordingly the degradation evaluation step (or means) may also include the identical first and second pitch conversion steps (or means), and the degradation evaluation step (or means) may further include an average signal difference calculation step of (or means) calculating an average signal difference by dividing a sum of power differences between a first pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the first pitch conversion step (or means) and a second pitch conversion result obtained by converting a part of the input signal pitch pattern per predetermined processing unit and the target pitch pattern at the second pitch conversion step (or means) per predetermined cycle by a sum of powers of the second pitch conversion result per predetermined cycle, and the degradation degree calculation step (or means) may provide as the degradation degree to the pitch conversion step (or means) a sum of values that are the average pitch conversion amount, the extracted degradation degree, and the average signal difference respectively weighted by predetermined coefficients.

[15] Also, in the above-mentioned [11], the degradation evaluation step (or means) may further include a pitch pattern change degree calculation step of (or means) classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and a degradation degree calculation step (or means) may provide as the degradation degree to the pitch conversion step (or means) a sum of values that are the average pitch conversion amount, the extracted degradation degree, and the pitch pattern change degree respectively weighted by predetermined coefficients.

[16] Also, in the above-mentioned [12], the degradation evaluation step (or means) may further include a pitch pattern change degree calculation step of (or means) classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and a degradation degree calculation step (or means) may provide as the degradation degree to the pitch conversion step (or means) a sum of values that are the average signal difference, the extracted degradation degree, and the pitch pattern change degree respectively weighted by predetermined coefficients.

[17] Also, in the above-mentioned [14], the degradation evaluation step (or means) may further include a pitch pattern change degree calculation step of (or means) classifying changing trends of the input signal pitch pattern and the target pitch pattern respectively into any one of predetermined changing trends by calculating average pitches per predetermined time interval of the pitch pattern and by sequentially comparing the average pitches, and determining a pitch pattern change degree to the target pitch pattern for the input signal pitch pattern based on a combination of both changing trends, and a degradation degree calculation step (or means) may provide as the degradation degree to the pitch conversion step (or means) a sum of values that are the average pitch conversion amount, the extracted degradation degree, the average signal difference, and the pitch pattern change degree respectively weighted by predetermined coefficients.

As the above-mentioned [11]-[17], the combination of two, three, or four of the average pitch conversion amount, the average signal difference, the pitch pattern change degree, and the degradation degree extracted at the degradation degree extraction step can be used as the degradation degree.

According to the present invention, the data throughput can be reduced while the degradation of the sound quality due to the pitch conversion can be suppressed as much as possible, thereby enabling a processing congestion of a device to which the present invention is applied and a delay of the pitch conversion due to the congestion to be prevented. Also, a long-lived device can be realized.

Also, it is made possible to easily calculate or extract the degradation degree, so that circuits within the device can be simplified.

BRIEF DESCRIPTION OF THE DRAWINGS

The above and other objects and advantages of the invention will be apparent upon consideration of the following detailed description, taken in conjunction with the accompanying drawings, in which the reference numerals refer to like parts throughout and in which:

FIG. 1 is a block diagram showing an embodiment [1] of a pitch conversion method and device according to the present invention;

FIG. 2 is a flowchart showing an entire operation example of a pitch conversion method and device according to the present invention;

FIG. 3 is a block diagram showing an embodiment (1) of a degradation evaluating portion used for an embodiment [1] of the present invention;

FIG. 4A is a flowchart showing an operation example (1) of a degradation evaluating portion used for an embodiment [1] of the present invention;

FIG. 4B is a temporal transition graph of an input signal pitch pattern and a target pitch pattern used for the present invention;

FIG. 5 is a block diagram showing an embodiment (2) of a degradation evaluating portion used for an embodiment [1] of the present invention;

FIG. 6 is a flowchart showing an operation example (2) of a degradation evaluating portion used for an embodiment [1] of the present invention;

FIG. 7 is a block diagram showing an embodiment (3) of a degradation evaluating portion used for an embodiment [1] of the present invention;

FIG. 8 is a flowchart showing an operation example (3) of a degradation evaluating portion used for an embodiment [1] of the present invention;

FIGS. 9A and 9B are diagrams showing examples of a pitch pattern change trend and a pitch pattern change degree calculating table used for the present invention;

FIGS. 10A and 10B are block diagrams showing an embodiment (4) of a degradation evaluating portion used for the embodiment [1] of the present invention;

FIGS. 11A and 11B are block diagrams showing an embodiment (5) of a degradation evaluating portion used for the embodiment [1] of the present invention;

FIGS. 12A and 12B are block diagrams showing an embodiment (6) of a degradation evaluating portion used for the embodiment [1] of the present invention;

FIGS. 13A and 13B are block diagrams showing an embodiment (7) of a degradation evaluating portion used for the embodiment [1] of the present invention;

FIG. 14 is a block diagram showing an embodiment [2] of a pitch conversion method and device according to the present invention;

FIG. 15A is a flowchart showing an operation example of a degradation degree extractor;

FIG. 15B is a diagram showing an example of a degradation rule database used for an embodiment [2] of the present invention;

FIG. 16 is a block diagram showing an embodiment [3] of a pitch conversion method and device according to the present invention;

FIGS. 17A and 17B are block diagrams showing an embodiment (8) of a degradation evaluating portion used for an embodiment [3] of the present invention;

FIGS. 18A and 18B are block diagrams showing an embodiment (9) of a degradation evaluating portion used for an embodiment [3] of the present invention;

FIGS. 19A and 19B are block diagrams showing an embodiment (10) of a degradation evaluating portion used for an embodiment [3] of the present invention;

FIGS. 20A and 20B are block diagrams showing an embodiment (11) of a degradation evaluating portion used for an embodiment [3] of the present invention;

FIGS. 21A and 21B are block diagrams showing an embodiment (12) of a degradation evaluating portion used for an embodiment [3] of the present invention;

FIGS. 22A and 22B are block diagrams showing an embodiment (13) of a degradation evaluating portion used for an embodiment [3] of the present invention;

FIGS. 23A and 23B are block diagrams showing an embodiment (14) of a degradation evaluating portion used for an embodiment [3] of the present invention; and

FIG. 24 is a time chart showing a prior art example [1] of a pitch conversion technology.

DESCRIPTION OF THE EMBODIMENTS

Embodiments [1]-[3] of a pitch conversion method and a device using the method according to the present invention will now be described in the following order by referring to FIGS. 1-23A, 23 B.

I. Embodiment [1]: FIGS. 1-13A, 13 B

I.1. Arrangement (common to embodiments (1)-(7) of degradation evaluating portion): FIG. 1

I.2. Entire operation example (common to embodiments [2] and [3]): FIG. 2

• • I.3. Embodiments (1)-(7) of degradation evaluating portion: FIGS. 3-13A, 13 B
    • I.3.A Embodiment (1) of degradation evaluating portion: FIGS. 3, 4A, and 4 B
    • I.3.A.a Arrangement: FIG. 3
    • I.3.A.b Operation example: FIGS. 4A and 4B
  • I.3.B Embodiment (2) of degradation evaluating portion: FIGS. 5 and 6
    • I.3.B.a Arrangement: FIG. 5
    • I.3.B.b Operation example: FIG. 6
  • I.3.C Embodiment (3) of degradation evaluating portion: FIGS. 7-9A, 9 B
    • I.3.C.a Arrangement: FIG. 7
    • I.3.C.b Operation example: FIGS. 8, 9A, and 9 B
  • I.3.D Embodiment (4) of degradation evaluating portion: FIGS. 10A and 10B
  • I.3.E Embodiment (5) of degradation evaluating portion: FIGS. 11A and 11B
  • I.3.F Embodiment (6) of degradation evaluating portion: FIGS. 12A and 12B
  • I.3.G Embodiment (7) of degradation evaluating portion: FIGS. 13A and 13 B
    II. Embodiment [2]: FIGS. 14, 15A, and 15 B

II. 1. Arrangement: FIG. 14

II. 2. Operation example: FIGS. 15A and 15B

III. Embodiment [3]: FIGS. 16-23A, 23 B

III.1. Arrangement (common to embodiments (8)-(14) of degradation evaluating portion): FIG. 16

III.2. Operation example: FIGS. 17 A, 17 B- 23 A, 23 B

III.3. Embodiments (8)-(14) of degradation evaluating portion: FIGS. 17 A, 17 B- 23 A, 23 B

• • III.3.A Embodiment (8) of degradation evaluating portion: FIGS. 17A and 17B
  • III.3.B Embodiment (9) of degradation evaluating portion: FIGS. 18A and 18B
  • III.3.C Embodiment (10) of degradation evaluating portion: FIGS. 19A and 19B
  • III.3.D Embodiment (11) of degradation evaluating portion: FIGS. 20A and 20B
  • III.3.E Embodiment (12) of degradation evaluating portion: FIGS. 21A and 21B
  • III.3.F Embodiment (13) of degradation evaluating portion: FIGS. 22A and 22B
  • III.3.G Embodiment (14) of degradation evaluating portion: FIGS. 23A and 23B

I. Embodiment [1]

FIGS. 1 - 13 A, 13 B

I.1. Arrangement (Common to Embodiments (1)-(7) of Degradation Evaluating Portion): FIG. 1

A pitch conversion device 10 according to an embodiment [1] of the present invention shown in FIG. 1 is composed of a degradation evaluating portion 100 which receives an input signal pitch pattern IPP per predetermined processing unit, a target pitch pattern TPP for the pitch pattern IPP, and a pitch mark PM to calculate a degradation degree DGR, and a pitch converter 200 which performs a pitch conversion depending on the degradation degree DGR.

The pitch mark PM is data indicating positions of pitch cycles (periods) within the input signal pitch pattern IPP and the target pitch pattern TPP. Also, a predetermined processing unit is a data unit of e.g. a predetermined number of pitch cycles (namely, a predetermined number of pitch marks PM), a single phoneme, a single voice fragment (assembly of a plurality of phonemes), a single sentence, or the like.

Also, the pitch converter 200 is composed of a pitch converter 310 (i.e. a low-performance pitch converter using the pitch conversion technology such as the above-mentioned prior art example [1]) which receives the input signal pitch pattern IPP, the target pitch pattern TPP, and the pitch mark PM to execute the pitch conversion with small data throughput, a pitch converter 320 (i.e. a high-performance pitch converter using a pitch conversion technology such as mentioned in the above-mentioned prior art example [2]) which executes the pitch conversion with large data throughput, and a switchover portion 400 which determines whether the pitch conversion should be performed either by the pitch converter 310 or 320 and switches over from one to the other.

Hereinafter, the operation of this embodiment will be described. An entire operation example will be firstly described referring to FIG. 2. Then, embodiments (1)-(7) of the degradation evaluating portion 100 will be described referring to FIGS. 3-13A, 13 B.

It is to be noted that the following description of the entire operation example is similarly applied to the embodiments [2] and [3] which will be described later except the calculation or extraction of the degradation degree DGR (hereinafter, referred to as degradation evaluation).

I.2. Entire Operation (Common to Embodiments [2] and [3]): FIG. 2

As shown in FIG. 2, the degradation evaluating portion 100 receives the input signal pitch pattern IPP per predetermined processing unit, the pitch mark PM, and the target pitch pattern TPP (at step S 1 ), and provides the degradation degree DGR obtained by executing the degradation evaluating which will be described later to the switchover portion 400 within the pitch converter 200 (at step S 2 ).

The switchover portion 400 compares the degradation degree DGR with a predetermined threshold “Th”. With the result determining that the degradation degree is less than the threshold “Th” (at step S 3 ), the switchover portion 400 provides the input signal pitch pattern IPP, the pitch mark PM, and the target pitch pattern TPP to the pitch converter 310 .

The pitch converter 310 having received the input signal pitch pattern IPP, the pitch mark PM, and the target pitch pattern TPP executes the pitch conversion (at step S 4 ), and transmits the output signal Out 1 after the pitch conversion to the subsequent stage (at step S 5 ).

On the other hand, with the result determining that the degradation degree is equal to or more than the threshold “Th” at the above-mentioned step S 3 , the switchover portion 400 provides the input signal pitch pattern IPP, the pitch mark PM, and the target pitch pattern TPP to the pitch converter 320 .

The pitch converter 320 having received the input signal pitch pattern IPP, the pitch mark PM, and the target pitch pattern TPP executes the pitch conversion (at step S 6 ), and transmits the output signal Out 2 after the pitch conversion to the subsequent stage (at step S 7 ).

I.3. Embodiments (1)-(7) of Degradation Evaluating Portion: FIGS. 3-13A, 13 B

I.3.A Embodiment (1) of Degradation Evaluating Portion: FIGS. 3, 4A, and 4 B

I.3.A.a Arrangement: FIG. 3

The degradation evaluating portion 100 shown in FIG. 3 is provided with an average pitch conversion amount calculator 110 which receives the input signal pitch pattern IPP, the pitch mark PM, and the target pitch pattern TPP to calculate an average pitch conversion amount PC, and a degradation degree calculator 120 which calculates the degradation degree DGR from the average pitch conversion amount PC.

I..3.A.b Operation Example: FIGS. 4A and 4B

As shown in FIG. 4A, the average pitch conversion amount calculator 110 calculates the average pitch conversion amount PC for the input signal according to the following equation (1) to be provided to the degradation degree calculator 120 (average pitch conversion amount calculation T 1 of step S 10 ). $\begin{array}{cc}·\mathrm{average}\mathrm{pitch}\mathrm{conversion}\mathrm{amount}\mathrm{PC}=\frac{\sum _{i=0}^n\Delta p_i}{\sum _{i=0}^n{\mathrm{IP}}_i}& \mathrm{Eq}.\left(1\right)\end{array}$

As shown in FIG. 4B, Δp _i in Eq. (1) indicates the absolute value of a pitch difference between a target pitch TP _i and an input signal pitch IP _i at the position of a pitch cycle shown by a pitch mark PM _i . The average pitch conversion amount PC is calculated by dividing the sum of the Δp _i (in the example of FIG. 4B, a pitch cycle number “n” per processing unit is assumed to be “10” (pitch cycles T 1 -T 10 )) by the sum of the input signal pitches IP _i .

The degradation degree calculator 120 calculates the degradation degree DGR by the following Eq. (2) based on the average pitch conversion amount PC to be provided to the switchover portion 400 (at step S 11 ). $\begin{array}{cc}\begin{array}{c}·\mathrm{degradation}\mathrm{degree}\mathrm{DGR}=f1\left(\mathrm{PC}\right)\\ =a·\mathrm{PC}+b\end{array}& \mathrm{Eq}.\left(2\right)\end{array}$

Coefficients “a” and “b” in the above-mentioned function f 1 have only to be preset by an operator or the like so that a switchover between the pitch converters 310 and 320 depending on the degradation degree DGR is optimally performed. The same applies to coefficients in functions used for embodiments of the degradation evaluating portion which will be described later.

I.3.B Embodiment (2) of Degradation Evaluating Portion: FIGS. 5 and 6

I.3.B.a Arrangement: FIG. 5

The degradation evaluating portion 100 shown in FIG. 5 is provided with an average signal difference calculator 130 which inputs a part of the input signal pitch pattern IPP, the pitch mark PM, and the target pitch pattern TPP to calculate an average signal difference DIF, and the degradation degree calculator 120 which calculates the degradation degree DGR from the average signal difference DIF.

Also, the average signal difference calculator 130 includes the pitch converters 310 and 320 which are the same as the pitch converters 310 and 320 shown in FIG. 1, and a signal difference calculator 131 which calculates the average signal difference DIF from the output signals Out 1 and Out 2 of the pitch converters 310 and 320 .

1.3.B.b Operation Example: FIG. 6

As shown in FIG. 6, the average signal difference calculator 130 executes an average signal difference calculation T 2 to calculate the average signal difference DIF of the output signal Out 1 from the output signal Out 2 .

Namely, the average signal difference calculator 130 inputs the input signal pitch pattern IPP, the pitch mark PM, and the target pitch pattern TPP for the pitch cycles “m” (smaller number than the pitch cycle number per processing unit) to be respectively provided to the pitch converters 310 and 320 (at step S 20 ).

The pitch converters 310 and 320 respectively execute the pitch conversion, and provides the output signals Out 1 and Out 2 after the pitch conversion to the signal difference calculator 131 (at steps S 21 and S 22 ).

The signal difference calculator 131 having received the output signals Out 1 and Out 2 calculates the average signal difference DIF according to the following Eq. (3) to be provided to the degradation degree calculator 120 (at step S 23 ). $\begin{array}{cc}·\mathrm{average}\mathrm{signal}\mathrm{difference}\mathrm{DIF}=\frac{\sum _{i=0}^m{\left(\mathrm{Out}1i-\mathrm{Out}2i\right)}^2}{\sum _{i=0}^m\mathrm{Out}2i^2}& \mathrm{Eq}.\left(3\right)\end{array}$

Out 1 _i and Out 2 _i in Eq. (3) indicate pitch conversion results obtained by the pitch conversion to an input signal pitch and a target pitch at the position of the pitch cycle shown by a pitch mark PM _i (see FIG. 4B) by the pitch converters 310 and 320 respectively. By dividing the sum of the power difference between the pitch conversion results Out 1 _i and Out 2 _i by the sum of powers of the pitch conversion results Out 2 _i , the average signal difference DIF is calculated.

The degradation degree calculator 120 calculates the degradation degree DGR by the following Eq. (4) based on the average signal difference DIF to be provided to the switchover portion 400 (at step S 24 ). $\begin{array}{cc}\begin{array}{c}·\mathrm{degradation}\mathrm{degree}\mathrm{DGR}=f2\left(\mathrm{DIF}\right)\\ =c·\mathrm{DIF}+d\\ \left(“c”\mathrm{and}“d”\mathrm{are}\mathrm{coefficients}\right)\end{array}& \mathrm{Eq}.\left(4\right)\end{array}$
I.3.C Embodiments (3) of Degradation Evaluating Portion: FIGS. 7-9A, 9 B
I.3.C.a Arrangement: FIG. 7

The degradation evaluating portion 100 shown in FIG. 7 is provided with a pitch pattern change degree calculating table TBL in which a change trend that the input signal pitch pattern IPP and the target pitch pattern TPP may transition is associated with a pitch pattern change degree CHG to be recorded, a pitch pattern change degree calculator 140 which receives the input signal pitch pattern IPP, the pitch mark PM, and the target pitch pattern TPP, and determines the pitch pattern change degree CHG by referring to the table TBL to be outputted, and the degradation degree calculator 120 which calculates the degradation degree DGR from the pitch pattern change degree CHG.

I.3.C.b Operation Example: FIGS. 8, 9A, and 9 B

As shown in FIG. 8, the pitch pattern change degree calculator 140 executes a pitch pattern change degree calculation T 3 to determine the pitch pattern change degree CHG to the target pitch pattern TPP with respect to the input signal pitch pattern IPP.

Namely, the pitch pattern change degree calculator 140 receives the input signal pitch pattern IPP, the pitch mark PM, and the target pitch pattern TPP (at step S 30 ), and calculates a change trend TND_I of the input signal pitch pattern IPP and a change trend TND_T of the target pitch pattern TPP (hereinafter, occasionally represented by a reference character TND) (at steps S 31 and S 32 ).

The pitch pattern change degree calculator 140 calculates average pitches AP 1 -AP 3 (hereinafter, occasionally represented by a reference character AP) for three predetermined time intervals of the pitch pattern (e.g. time that is a pitch cycle divided into three, shown by the pitch mark PM), as shown in FIG. 9A, sequentially compares the average pitches AP 1 -AP 3 , and classifies the pitch pattern change trends TND into any one of nine pitch pattern change trends TND 1 -TND 9 .

If the average pitches AP 1 -AP 3 of the input signal pitch pattern satisfy the relationship of AP 1 <AP 2 <AP 3 (namely, a change trend that the average pitch AP gradually increases) for example, the pitch pattern change degree calculator 140 classifies the input signal pitch pattern change trend TND_I into a pitch pattern change trend TND 1 .

The pitch pattern change degree calculator 140 determines the pitch pattern change degree CHG from the combination of the input signal pitch pattern change trend TND_I and the target pitch pattern change trend TND_T by referring to the pitch pattern change degree calculating table TBL shown in FIG. 9B (at step S 33 ).

As shown, the pitch pattern change degree calculating table TBL is set so that as the difference between the input signal pitch pattern change trend TND_I and the target pitch pattern change degree TND_T becomes large, a larger value is obtained as the pitch pattern change degree CHG.

When the input signal pitch pattern change trend TND_I and the target pitch pattern change trend TND_T are respectively classified into a pitch pattern change trend TND 3 (change trend in which the average pitch AP changes from up to down) and a pitch pattern change trend TND 7 (change trend in which the average pitch AP changes from down to up) (namely, when the difference of the pitch pattern change trend TND is the largest) for example, the pitch pattern change degree calculator 140 determines the pitch pattern change degree CHG to be “4” (maximum value) by referring to the pitch pattern change degree calculating table TBL.

The degradation degree calculator 120 calculates the degradation degree DGR by the following Eq. (5) based on the pitch pattern change degree CHG to be provided to the switchover portion 400 (at step S 34 ).
degradation degree DGR=f3( CHG ) Eq. (5)

For the above-mentioned function f 3 , the same function as the function f 1 or f 2 described in the above-mentioned embodiment (1) or (2) of the degradation evaluating portion can be used.

I..3.D Embodiment (4) of Degradation Evaluating Portion: FIGS. 10A and 10B

The degradation evaluating portion 100 shown in FIG. 10A is provided with the average signal difference calculator 130 which is the same as that of the above-mentioned embodiment (2) of the degradation evaluating portion, in addition to the arrangement of the above-mentioned embodiment (1) of the degradation evaluating portion. However, this embodiment is different from the embodiment (2) in that the degradation degree calculator 120 calculates the degradation degree DGR from the average pitch conversion amount PC and the average signal difference DIF respectively provided from the average pitch conversion amount calculator 110 and the average signal difference calculator 130 .

In operation, as shown in FIG. 10B, the average pitch conversion amount calculator 110 and the average signal difference calculator 130 respectively execute the above-mentioned average pitch conversion amount calculation and average signal difference calculation to calculate the average pitch conversion amount PC and the average signal difference DIF (at steps T 1 and T 2 ).

The degradation degree calculator 120 calculates the degradation degree DGR by the following Eq. (6) based on the average pitch conversion amount PC and the average signal difference DIF to be provided to the switchover portion 400 (at step S 40 ). $\begin{array}{cc}\begin{array}{c}·\mathrm{degradation}\mathrm{degree}\mathrm{DGR}=f4\left(\mathrm{PC},\mathrm{DIF}\right)\\ =\alpha 1·f1\left(\mathrm{PC}\right)+\left(1-\alpha 1\right)·\\ f2\left(\mathrm{DIF}\right)\left(\alpha 1\mathrm{is}\mathrm{coefficient}\right)\end{array}& \mathrm{Eq}.\left(6\right)\end{array}$
I.3.E Embodiment (5) of Degradation Evaluating Portion: FIGS. 11A and 11B

The degradation evaluating portion 100 shown in FIG. 11A is provided with the pitch pattern change degree calculator 140 and the pitch pattern change degree calculating table TBL which are the same as those of the above-mentioned embodiment (3) of the degradation evaluating portion, in addition to the arrangement of the above-mentioned embodiment (1) of the degradation evaluating portion. However, this embodiment is different from the embodiment (3) in that the degradation degree calculator 120 calculates the degradation degree DGR based on the average pitch conversion amount PC and the pitch pattern change degree CHG respectively provided from the average pitch conversion amount calculator 110 and the pitch pattern change degree calculator 140 .

In operation, as shown in FIG. 11B, the average pitch conversion amount calculator 110 and the pitch pattern change degree calculator 140 respectively execute the above-mentioned average pitch conversion amount calculation and pitch pattern change degree calculation to calculate the average pitch conversion amount PC and the pitch pattern change degree CHG (at steps T 1 and T 3 ).

The degradation degree calculator 120 calculates the degradation degree DGR by the following Eq. (7) based on the average pitch conversion amount PC and the pitch pattern change degree CHG to be provided to the switchover portion 400 (at step S 50 ). $\begin{array}{cc}\begin{array}{c}·\mathrm{degradation}\mathrm{degree}\mathrm{DGR}=f5\left(\mathrm{PC},\mathrm{CHG}\right)\\ =\alpha 2·f1\left(\mathrm{PC}\right)+\left(1-\alpha 2\right)·\\ f3\left(\mathrm{CHG}\right)\left(\alpha 2\mathrm{is}\mathrm{coefficient}\right)\end{array}& \mathrm{Eq}.\left(7\right)\end{array}$
I.3.F Embodiment (6) of Degradation Evaluating Portion: FIGS. 12A and 12B

The degradation evaluating portion 100 shown in FIG. 12A is provided with the pitch pattern change degree calculator 140 and the pitch pattern change degree calculating table TBL which are the same as those of the above-mentioned embodiment (3) of the degradation evaluating portion, in addition to the arrangement of the above-mentioned embodiment (2) of the degradation evaluating portion. However, this embodiment is different from the embodiment (3) in that the degradation degree calculator 120 calculates the degradation degree DGR based on the average signal difference DIF and the pitch pattern change degree CHG respectively provided from the average signal difference calculator 130 and the pitch pattern change degree calculator 140 .

In operation, as shown in FIG. 12B, the average signal difference calculator 130 and the pitch pattern change degree calculator 140 respectively execute the above-mentioned average signal difference calculation and pitch pattern change degree calculation to calculate the average signal difference DIF and the pitch pattern change degree CHG (at steps T 2 and T 3 ).

The degradation degree calculator 120 calculates the degradation degree DGR by the following Eq. (8) based on the average signal difference DIF and the pitch pattern change degree CHG to be provided to the switchover portion 400 (at step S 60 ). $\begin{array}{cc}\begin{array}{c}·\mathrm{degradation}\mathrm{degree}\mathrm{DGR}=f6\left(\mathrm{DIF},\mathrm{CHG}\right)\\ =\alpha 3·f2\left(\mathrm{DIF}\right)+\left(1-\alpha 3\right)·\\ f3\left(\mathrm{CHG}\right)\left(\alpha 3\mathrm{is}\mathrm{coefficient}\right)\end{array}& \mathrm{Eq}.\left(8\right)\end{array}$
I.3.G Embodiment (7) of Degradation Evaluating Portion: FIGS. 13A and 13B

The degradation evaluating portion 100 shown in FIG. 13A is provided with the pitch pattern change degree calculator 140 and the pitch pattern change degree calculating table TBL which are the same as those of the above-mentioned embodiment (3) of the degradation evaluating portion, in addition to the arrangement of the above-mentioned embodiment (4) of the degradation evaluating portion. However, this embodiment is different from the embodiment (3) in that the degradation degree calculator 120 calculates the degradation degree DGR based on the average pitch conversion amount PC, the average signal difference DIF, and the pitch pattern change degree CHG respectively provided from the average pitch conversion amount calculator 110 , the average signal difference calculator 130 , and the pitch pattern change degree calculator 140 .

In operation, as shown in FIG. 13B, the average pitch conversion amount calculator 110 , the average signal difference calculator 130 , and the pitch pattern change degree calculator 140 respectively execute the above-mentioned average pitch conversion amount calculation, average signal difference calculation, and pitch pattern change degree calculation to respectively calculate the average pitch conversion amount PC, the average signal difference DIF, and the pitch pattern change degree CHG (at steps T 1 -T 3 ).

The degradation degree calculator 120 calculates the degradation degree DGR by the following Eq. (9) based on the average pitch conversion amount PC, the average signal difference DIF, and the pitch pattern change degree CHG to be provided to the switchover portion 400 (at step S 70 ). $\begin{array}{cc}\begin{array}{c}·\mathrm{degradation}\mathrm{degree}\mathrm{DGR}=f7\left(\mathrm{PC},\mathrm{DIF},\mathrm{CHG}\right)\\ =\beta 1·f1\left(\mathrm{PC}\right)+\beta 2·f2\left(\mathrm{DIF}\right)+\\ \beta 3·f3\left(\mathrm{CHG}\right)\end{array}\left(\beta 1·\beta 3\mathrm{are}\mathrm{coefficients}\mathrm{satisfying}\beta 1+\beta 2+\mathrm{\beta 3}=1\right)& \mathrm{Eq}.\left(9\right)\end{array}$

II. Embodiment [2]

FIGS. 14 , 15 A, and 15 B

II.1. Arrangement: FIG. 14

The pitch conversion device 10 according to the embodiment [2] of the present invention shown in FIG. 14 is arranged so as to include, substituting for the degradation evaluating portion 100 in the above-mentioned embodiment [1], a degradation rule database DB in which a combination of all of the voice states and phonemic types estimated as the input signal are associated with the degradation degree DGR to be recorded, and a degradation degree extractor 500 which receives additional information INFO indicating the sound state and the phonemic type of the input signal to extract the degradation degree DGR from the database DB.

The sound state of the additional information INFO indicates a state such as “rise”, “fall, “transition”, and “steady” estimated as the input signal, and the phonemic type indicates a type such as vowels (“A”-“O”) and consonants (except vowels). The relationship between all of the combinations of the voice states and the phonemic types, and the degradation degree DGR (namely, degradation of sound quality which may actually occur) is preliminarily obtained by a simulation, an experiment, or the like to be recorded in the degradation rule database DB.

Hereinafter, the operation of this embodiment will be described. However, since operations except extraction of the degradation degree DGR in the degradation degree extractor 500 is common to that of the above-mentioned embodiment [1], only the operation of the degradation degree extractor 500 will now be described referring to FIGS. 15A and 15B.

II.2. Operation Example: FIGS. 15A and 15B

As shown in FIG. 15A, the degradation degree extractor 500 extracts the degradation degree DGR corresponding to the voice state and the phonemic type indicated by the inputted additional information INFO from the degradation rule database DB shown in FIG. 15B to be provided to the switchover portion 400 (degradation degree extraction T 4 ).

When the voice state and the phonemic type of the additional information INFO respectively indicate the “transition” state and the vowel “O” for example, the degradation degree extractor 500 extracts “10” for the degradation degree DGR from the degradation rule database DB.

III. Embodiment [3]

FIGS. 16 - 23 A, 23 B

III.1. Arrangement (Common to Embodiments (8)-(14) of Degradation Evaluating Portion): FIG. 16

The pitch conversion device 10 according to the embodiment [3] of the present invention shown in FIG. 16 is arranged so that the additional information INFO indicating the voice state and the phonemic type of the input signal is inputted to the degradation evaluating portion 100 in addition to the arrangement of the above-mentioned embodiment [1].

III.2. Operation Example: FIGS. 17 A, 17 B- 23 A, 23 B

While the operation of this embodiment will be described hereinafter, only the embodiments (8)-(14) of the degradation evaluating portion 100 will now be described referring to FIGS. 17 A, 17 B- 23 A, 23 B since the arrangement and the operation except the calculation of the degradation degree DGR in the degradation evaluating portion 100 are the same as those in the above-mentioned embodiments [1] and [2].

III.3. Embodiments (8)-(14) of Degradation Evaluating Portion: FIGS. 17 A, 17 B- 23 A, 23 B

III.3.A Embodiment (8) of Degradation Evaluating Portion: FIGS. 17A and 17B

In addition to the average pitch conversion amount calculator 110 , the degradation degree extractor 500 , and the degradation rule database DB which are the same as those of the above-mentioned embodiments [1] and [2], the degradation evaluating portion 100 shown in FIG. 17A is provided with the degradation degree calculator 120 which calculates the degradation degree DGR based on the average pitch conversion amount PC and the degradation degree DGR respectively provided from the calculator 110 and the extractor 500 .

In operation, as show in FIG. 17B, the average pitch conversion amount calculator 110 and the degradation degree extractor 500 respectively execute the above-mentioned average pitch conversion amount calculation and degradation degree extraction to calculate the average pitch conversion amount PC and to extract the degradation degree DGR (at steps T 1 and T 4 ).

The degradation degree calculator 120 calculates the degradation degree DGR by the following Eq. (10) based on the average pitch conversion amount PC and the degradation degree DGR to be provided to the switchover portion 400 (at step S 80 ). $\begin{array}{cc}\begin{array}{c}·\mathrm{degradation}\mathrm{degree}\mathrm{DGR}=f8\left(\mathrm{PC},\mathrm{DGR}\right)\\ =\alpha 4·f1\left(\mathrm{PC}\right)+\left(1-\alpha 4\right)·\\ \mathrm{DGR}\end{array}& \mathrm{Eq}.\left(10\right)\end{array}$

The coefficient α4 in the above-mentioned function f 8 may be preset by an operator or the like so that the switchover between the pitch converters 310 and 320 depending on the degradation degree DGR is optimally performed in the same way as the above-mentioned embodiment [1]. The same applies to coefficients in functions used for embodiments of the degradation evaluating portion as will be described later.

III.3.B Embodiment (9) of Degradation Evaluating Portion: FIGS. 18A and 18B

In addition to the average signal difference calculator 130 , the degradation degree extractor 500 , and the degradation rule database DB which are the same as those of the above-mentioned embodiments [1] and [2], the degradation evaluating portion 100 shown in FIG. 18A is provided with the degradation degree calculator 120 which calculates the degradation degree DGR based on the average signal difference DIF and the degradation degree DGR respectively outputted from the calculator 130 and the extractor 500 .

In operation, as show in FIG. 18B, the average signal difference calculator 130 and the degradation degree extractor 500 respectively execute the above-mentioned average signal difference calculation and degradation degree extraction to calculate the average signal difference DIF and to extract the degradation degree DGR (at steps T 2 and T 4 ), respectively.

The degradation degree calculator 120 calculates the degradation degree DGR by the following Eq. (11) based on the average signal difference DIF and the degradation degree DGR to be provided to the switchover portion 400 (at step S 90 ). $\begin{array}{cc}\begin{array}{c}·\mathrm{degradation}\mathrm{degree}\mathrm{DGR}=f9\left(\mathrm{PC},\mathrm{DGR}\right)\\ =\alpha 5·f2\left(\mathrm{DIF}\right)+\left(1-\alpha 5\right)·\\ \mathrm{DGR}\left(\alpha 5\mathrm{is}\mathrm{coefficient}\right)\end{array}& \mathrm{Eq}.\left(11\right)\end{array}$
III.3.C Embodiment (10) of Degradation Evaluating Portion: FIGS. 19A and 19B

In addition to the pitch pattern change degree calculator 140 , the pitch pattern change degree calculating table TBL, the degradation degree extractor 500 , and the degradation rule database DB which are the same as those in the above-mentioned embodiments [1] and [2], the degradation evaluating portion 100 shown in FIG. 19A is provided with the degradation degree calculator 120 which calculates the degradation degree DGR based on the pitch pattern change degree CHG and the degradation degree DGR respectively provided from the calculator 150 and the extractor 500 .

In operation, as shown in FIG. 19B, the pitch pattern change degree calculator 140 and the degradation degree extractor 500 respectively execute the above-mentioned pitch pattern change degree calculation and degradation degree extraction to calculate the pitch pattern change degree CHG and to extract the degradation degree DGR (at steps T 3 and T 4 ), respectively.

The degradation degree calculator 120 calculates the degradation degree DGR by the following Eq. (12) based on the pitch pattern change degree CHG and the degradation degree DGR to be provided to the switchover portion 400 (at step S 100 ). $\begin{array}{c}\begin{array}{c}\mathrm{degradation}\mathrm{degree}DGR=f10\\ \left(CHG,DGR\right)\end{array}\end{array}$