Title:
AUDIO ENHANCEMENT SYSTEM AND METHOD
Kind Code:
A1


Abstract:
A system and method for enhancing the sound signal produced by an audio system in a listening environment by compensating for ambient sound in the listening environment, comprises producing an audio sound in the time domain from an electrical sound signal in the time domain. The electrical sound signal in the time domain is transformed into an electrical sound signal in the frequency domain and the electrical sound signal in the frequency domain is retransformed into an audio sound in the time domain. The total sound level in the environment is measured and a signal representative thereof is generated. The audio sound signal and the total sound signal are processed to extract a signal representing the ambient sound level within the environment, and equalization is performed in the frequency domain to adjust the output from the audio sound signal to compensate for the ambient noise level.



Inventors:
Christoph, Markus (Straubing, DE)
Application Number:
12/246101
Publication Date:
02/05/2009
Filing Date:
10/06/2008
Primary Class:
International Classes:
G10K11/04
View Patent Images:
Related US Applications:
20090092273Silicon microphone with enhanced impact proof structure using bonding wiresApril, 2009Zhe et al.
20080192977RADIO WITH DUAL SIDED AUDIOAugust, 2008Gruenhagen et al.
20050283263Hearing aid systemsDecember, 2005Eaton et al.
20060062404Cross-over frequency selection and optimization of response around cross-overMarch, 2006Bharitkar et al.
20080044047Identification element for a hearing device unitFebruary, 2008Kral et al.
20020048375Circuit housing for a headsetApril, 2002Bebenroth
20080089528Sound altering apparatusApril, 2008Klossowski
20090103741Method of correction of acoustic parameters of electro-acoustic transducers and device for its realizationApril, 2009Skuruls
20070189550Panel acoustic contributions examinationAugust, 2007Wu
20090232327AUTOMOTIVE SENSORY ENHANCEMENT SYSTEMSeptember, 2009Hagen
20090088224ADAPTIVE VOLUME CONTROLApril, 2009Le Gall et al.



Primary Examiner:
ORTIZ SANCHEZ, MICHAEL
Attorney, Agent or Firm:
Patrick J. O'Shea, Esq. (Springfield, MA, US)
Claims:
What is claimed is:

1. An audio enhancement system for compensating for ambient noise in a listening environment, comprising: means for producing an electrical sound signal in the time domain and generating a sound output from the electrical sound signal, comprising time-to-frequency transformation means responsive to the electrical sound signal in the time domain for generating an electrical sound signal in the frequency domain; frequency-to-time transformation means responsive to the electrical sound signal in the frequency domain for generating a re-transformed electrical sound signal in the time domain for generating the sound output; a sensor that obtains a sound signal representative of the total sound level in the environment, where the total sound level comprises both the sound output from the means for producing and the ambient noise within the environment; means, responsive to the total sound signal and to the electrical sound signal, for extracting an ambient noise signal representative of the ambient noise in the environment from the total sound signal, where spectral voice signal components in the total sound signal are suppressed; and means, responsive to the ambient noise signal and the electrical sound signal in the frequency domain, for equalizing according to the spectral distribution of the noise signal and adjusting in the frequency domain the sound output of the means for producing to compensate for the ambient noise level.

2. The audio enhancement system of claim 1, where the means for equalizing and adjusting comprises a gain shaping unit for performing an equalizing function in the frequency domain.

3. The audio enhancement system of claim 2, where the means for equalizing and adjusting comprises a voice activity detector for detecting and suppressing spectral voice components in the total sound signal.

4. The audio enhancement system of claim 3, where the means for equalizing and adjusting comprises a mean value calculation unit.

5. The audio enhancement system of claim 4, where the means for equalizing and adjusting comprises a time-to-frequency transformation unit connected to the sensor for transforming the total sound signal in the time domain into a total sound signal in the frequency domain.

6. The audio enhancement system of claims 5, where the means for extracting comprises a noise estimation unit operated in the frequency domain.

7. The audio enhancement system of claim 6, where the noise estimation unit comprises a frequency domain adaptive filter.

8. The audio enhancement system of claims 7, where the frequency domain adaptive filter comprises an overlap-save frequency domain adaptive filter.

9. The audio enhancement system of claims 5, where the means for extracting comprises a noise estimation unit operated in the time domain, the noise estimation unit being connected between the sensor and the means for equalizing and adjusting.

10. The audio enhancement system of claim 9, where the noise estimation unit operated in the time domain comprises at least one adaptive filter.

11. The audio enhancement system of claim 1, where the means for equalizing and adjusting performs a linear predictive coding in the frequency domain.

12. The audio enhancement system of claim 2, where the gain shaping unit performs a psycho-acoustically motivated gain shaping function.

13. The audio enhancement system of claim 12, where the gain shaping unit has a frequency scale according to psycho-acoustic frequency groups.

14. The audio enhancement system of claim 12, where the gain shaping unit has a warped frequency scale.

15. The audio enhancement system of claim 12, where the gain shaping unit performs psycho-acoustic masking.

16. The audio enhancement system of claim 15, where the psycho-acoustic masking is performed in the MPEG format.

17. The audio enhancement system of claim 15, where the psycho-acoustic masking is performed according to the Johnston model.

18. The audio enhancement system of claim 1, where the means for producing comprises a bypass path for higher frequencies, the bypass path having a controllable gain controlled by the means for equalizing and adjusting.

19. The audio enhancement system of claim 18, where the bypass path comprises a delay unit.

20. The audio enhancement system of claim 18, where the means for equalizing and adjusting is operated in an undersampled mode.

21. A method for enhancing the sound signal produced by an audio system in a listening environment by compensating for ambient noise in the listening environment, comprising the steps of: producing an audio sound in the time domain from an electrical sound signal in the time domain, the electrical sound signal in the time domain being transformed into electrical sound signal in the frequency domain and the electrical sound signal in the frequency domain being retransformed into audio sound in the time domain to provide an audio sound signal; measuring the sound level in the environment and generating a total sound signal representative thereof; processing the audio sound signal and the total sound signal to extract a signal representing the ambient noise level within the environment, where spectral voice signal components in the total sound signal are suppressed; and performing an equalization according to the spectral distribution of the signal representing the ambient noise level in the frequency domain and adjusting the output from the audio sound signal to compensate for the ambient noise level.

22. The method of claim 21, where equalizing comprises gain shaping in the frequency domain.

23. The method of claim 22, where equalizing comprises detecting a voice activity.

24. The method of claim 23, where equalizing comprises calculating a mean value.

25. The method of claim 21, where equalizing comprises a time-to-frequency transformation step for transforming the total sound signal in the time domain into a total sound signal in the frequency domain.

26. The method of claim 25, where the processing to extract comprises estimating noise estimation in the frequency domain.

27. The method of claim 26, where estimating the noise comprises adaptive filtering in the frequency domain.

28. The method of claim 27, where the adaptive filtering comprises an overlap-save adaptive filtering in the frequency domain.

29. The method of claim 25, where the extraction step comprises a noise estimation step in the time domain.

30. The method of claim 29, where the noise estimation step in the time domain comprises at least one adaptive filtering step.

31. The method of claim 30, where the equalizing comprises linear predictive coding in the frequency domain.

32. The method of claim 22, where the gain shaping step comprises psycho-acoustical gain shaping.

33. The method of claim 32, where the gain shaping step comprises frequency scaling according to psycho-acoustic frequency groups.

34. The method of claim 32, where the gain shaping step comprises a warped frequency scaling.

35. The method of claim 32, where the gain shaping step performs psycho-acoustic masking.

36. The method of claim 35, where the psycho-acoustic masking is performed in the MPEG format.

37. The method of claim 35, where the psycho-acoustic masking is performed according to the Johnston model.

38. The audio enhancement system of claim 1, where the means for producing comprises a bypass path for higher frequencies, the bypass path having a controllable gain controlled by the means for equalizing and adjusting.

39. The audio enhancement system of claim 38, where the bypass path comprises a delay step.

40. The audio enhancement system of claim 38, where the means for equalizing and adjusting operates in an undersampled mode.

41. An audio enhancement system for compensating for ambient noise in a listening environment, comprising: a time to frequency domain transformation unit that receives an input signal and provides a frequency domain output signal; a multiplier that receives the frequency domain output signal and an equalizing signal and provides a multiplier output signal; a frequency to time domain transformation unit that receives the multiplier output signal and provides an equalized signal; a loudspeaker that receives the equalized signal and outputs an audio signal indicative thereof to a listening environment; a transducer that senses audio within the listening environment and provides a sensed audio signal indicative thereof; means responsive to the sensed audio signal and the frequency domain output signal for providing the equalizing signal.

Description:

1. CLAIM OF PRIORITY

This application is a continuation of U.S. patent application Ser. No. 11/185,534 filed on Jul. 20, 2005, which is hereby incorporated by reference.

2. FIELD OF THE INVENTION

This invention relates to a system and method for improving the sound reproduced by audio systems in a listening environment such as a vehicle and, in particular, to a system and method which compensates for noise outside the audio system.

3. RELATED ART

When music or speech is reproduced, for example, in a vehicle, the signal is corrupted by external acoustic noise present in the vehicle. This noise may result from and is dependent upon vehicle speed, road condition, weather and condition of the vehicle. The presence of such noise results in a situation where soft sounds of interest are hidden, the perceived loudness is reduced and the intelligibility of the signal is decreased. The vehicle's driver and/or passengers may compensate for increased external noise by increasing the volume of the audio system. However, when the vehicle speed decreases or another source of external noise is alleviated, the volume of the audio system will be too high, requiring the user to reduce it. To overcome this, U.S. Pat. Nos. 5,434,922 and 6,529,605 propose an enhanced dynamic volume control (DVC) system which extracts the noise signal from a signal provided by a sensor (e.g., a sensing microphone) in the listening environment and calculates a control signal therefrom. This control signal is used to control the volume and/or dynamics of the desired signal (i.e., music).

A DVC system extracts the noise signal from mixed signals derived from a sensor (e.g., a microphone). The mixed signals comprise music components (i.e., the desired signal), voice components and noise components. The noise component is intended to be used solely for obtaining the control signal for the volume or dynamics of the desired signal. The other components are not wanted to have any effect on the derivation of the control signal as otherwise the system would respond to voice signals or control itself through the music, which would end in a so-called gain chase situation (i.e., direct feedback). Such a gain chase situation may lead to instability in the entire audio system.

The music signal is extracted from the sensor signal using an adaptive filter. The voice signal left in the remaining signal mixture is then masked out using a “voice activity detector” (VAD). The VAD operates continuously in the time domain—i.e., in a broadband manner—and is implemented by an averaging unit with permanently configured increments and decrements. With other words, as long as the input signal is larger than the output signal, the VAD increases its output signal by a fixed increment, or reduces it by a fixed decrement whenever the input signal is less than the output signal from the VAD. In this way, the VAD utilizes the different stationary properties of the voice and noise signals. The noise signal is strongly (broadband) smoothed so that the VAD output signal (i.e., the control signal), even if somewhat delayed, reaches a stationary final value, which approximately corresponds to the average power of the noise signal in the sensor signal. Depending on the configured volume or selected equalizing, which mainly refers in this instance to the bass setting, the music signal more or less penetrates the noise signal—i.e., the louder the desired signal (music signal) is played or the higher the bass controller is set, the greater the share of the music signal that passes unfiltered through the adaptive filter. This can lead to the known gain chase situation described above, which it is imperative to prevent.

It has been found that the adaptive filter works better (i.e., permits less of the desired signal to pass through it) if the signals have a narrower bandwidth. For this reason, the DVC system mostly works with strongly undersampled signals, which on the one hand reduces the implementation complexity, but on the other hand leads to a control signal derived solely from the low-frequency noise component and then applied in a broadband manner to control the volume or dynamics. Since low-frequency noise signals dominate in vehicles—the field for which most of the DVC system are primarily designed—the spectral limitation described above can only actually be considered in this context. Nonetheless, the solution is not ideal and may lead to overlapping effects in certain circumstances, which is why a broadband solution is preferable. Although the risk of gain chase is reduced through limiting the bandwidth, it is not fully eliminated.

One way of avoiding gain chase is to upward-limit the control signal in accordance with the existing power of the desired signal that is implemented in common systems in the form of an anti-gain chase function. This function permits the control signal, provided the desired signal is below a specific minimum threshold value, to pass through the filter without being changed, but limits it to a maximum value specified by another function if the power of the desired signal rises above the threshold, and blocks further control once the desired signal has exceeded a maximum threshold—i.e., the control signal is then replaced by zero. The control signal modified in this way can then be used to alter the volume and/or dynamics of the desired signal using a compressor. However, the control signal is, nevertheless, fully dependent on the mean power of the currently existing noise signal but does not consider its spectral distribution or coloring.

In this regard, systems known as “dynamic equalizer control” (DEC) systems are considered successors to DVC systems. However, one aspect that hinders the transition from DVC to DEC systems is the limited bandwidth with which DVC systems work. The reason why the bandwidth is limited is primarily to reduce the risk of gain chase and to reduce the implementation work.

Therefore, there is a need for an improved system and method that automatically compensate for the noise level in a listening environment.

SUMMARY OF THE INVENTION

A dynamic equalizer control (DEC) system digitally processes electrical audio signals to compensate for the intrusion of acoustic noise into a listening area. The system may be used with any audio system and any listening environment. For purposes of simplicity, it is referred to the output of the audio system as the music signal and the listening environment described herein is a vehicle cabin. It is understood that the invention is not so limited, as it may be used in other rooms and listening environments, as well.

The DEC system comprises an audio system that produces an electrical sound signal in the time domain and generates a sound output from the electrical sound signal. The audio system comprises a time-to-frequency transformation responsive to the electrical sound signal in the time domain for generating an electrical sound signal in the frequency domain. The audio system further comprising a frequency-to-time transform responsive to the electrical sound signal in the frequency domain for generating a re-transformed electrical sound signal in the time domain for generating the sound output. A sensor senses the sound representative of the total sound level in the environment, wherein the total sound level comprises both the sound output from the audio system and the ambient noise within the environment. An extraction circuit responsive to the total sound signal and to the electrical sound signal extracts an ambient noise signal representative of the ambient noise in the environment from the total sound signal. A controller responsive to the ambient noise signal and the electrical sound signal in the frequency domain for equalizes and adjusts in the frequency domain the sound output of the audio system to compensate for the ambient noise level.

A method in accordance with the present invention comprises producing an audio sound from an electrical sound signal in the time domain. The electrical sound signal in the time domain is transformed into electrical sound signal in the frequency domain and the electrical sound signal in the frequency domain is retransformed into audio sound in the time domain. The method further comprises measuring the total sound level in the environment and generating a signal representative thereof. In addition the audio sound signal and the total sound signal are processed to extract a signal representing the ambient sound level within the environment. Equalizing is performed in the frequency domain to adjust the output from the audio sound signal to compensate for the ambient noise level.

The DEC system measures the loudness of the music and noise in a vehicle and determines the effect of the noise on the perceived loudness of the music. The system then boosts the level of the music to compensate for the masking effect. The loudness of the music is determined by monitoring the music signal voltage coming from the music source. The sound in the vehicle includes both music and noise as measured by the microphone. The microphone signal is converted to a digital representation, and the system uses digital processing to remove the music signal. The system includes hardware and appropriate software to shape the spectrum of the music and noise signal to mimic human hearing.

An aspect of the present invention is that basic parts of the DEC system are operated in the frequency domain. The invention provides for a system that adjusts not only gain but equalization in response to the noise. The perception of high frequencies is not greatly affected by road noise, but bass frequencies are strongly masked. The DEC system controls the desired signal (e.g., music signal) according to the spectral distribution of the noise signal, and in doing so psycho-acoustic aspects may also be considered when configuring the equalization.

As can be seen, even though it is more complex to implement the DEC system in the spectral domain than in the time domain, the processing nevertheless benefits from greater flexibility and options, which in the final outcome represents superior quality of the system.

The other systems, methods, features and advantages of the invention will be, or will become, apparent to one with skill in the art upon examination of the following figures and detailed description. It is intended that all such additional systems, methods, features and advantages be included within this description, be within the scope of the invention, and be protected by the following claims.

DESCRIPTION OF THE DRAWINGS

The invention can be better understood with reference to the following drawings and description. The components in the figures are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the invention. Moreover, in the figures, like reference numerals designate corresponding parts throughout the different views.

FIG. 1 is a block diagram of a DEC system;

FIG. 2 is an alternative embodiment of a DEC system;

FIG. 3 is a frequency domain adaptive filter for a DEC system;

FIG. 4 illustrates the characteristics of masking thresholds compared to an absolute background threshold;

FIG. 5 is a diagram of the gain of a typical equalizer filter in an automobile;

FIG. 6 is a representation of the iterative computation of the IIR filter coefficients using the GAL algorithm;

FIG. 7 illustrates the comparison of the original equalizer filter with the approximated equalizer filter; and

FIG. 8 is an alternative implementation of a DEC system comprising a bypass.

DETAILED DESCRIPTION

Analysis shows that the main reason why adaptive filters fail in broadband applications in the time domain is to be found in the wide dynamic range of the input signals. If the broadband input signals in the time domain are transformed into the frequency domains in such a way that all frequency are processed, the music signal can be removed from the sensor signal in a broadband manner.

FIG. 1 illustrates a DEC system essentially operated in the frequency domain. In the system of FIG. 1, the voice signal component in the microphone signal is suppressed in the frequency domain. A signal source 101 supplies a desired signal, such as for example a music signal x[k] from CD player, radio, cassette player or the like. The signal x[k] is fed into a spectral dynamic equalization control (DEC) unit 102 operated in the frequency domain to an equalized signal y[k] on a line 107 to a loudspeaker 103. The loudspeaker 103 generates an acoustic signal, transferred to a microphone 104 wherein the transferal can be described by a transfer function H(z). The signal from the microphone 104 is supplied to an adaptive filter 106 operated in the time domain for estimating the noise in the microphone signal. The microphone provides a signal on a line 105 representing the total sound level in the environment, wherein the total sound level comprises both the sound output from the loudspeaker 103 and the ambient noise within the environment (e.g., the loudspeaker-room-microphone (LRM) system).

The signal y[k] on the line 107 from the DEC unit 102 is used as a reference signal for the adaptive filter 106. The signal output by the adaptive filter 106 is transferred via a spectral voice suppression unit 109 and a psycho-acoustic gain shaping unit 114 operated in the frequency domain to the DEC unit 102. The voice suppression unit 109 comprises a Fast-Fourier transform (FFT) unit 110 for transforming signals from the time domain into signals in the frequency domain. In a mean calculation unit 111 the signals in the frequency domain are averaged and supplied to a voice activity decoder (VAD) 112 for detecting and suppressing spectral voice signal components in the signals in the frequency domain from the FFT unit 110. Different kinds of voice activity detectors are known, for example, from US20030053639A1.

The signals from the VAD 112 are supplied to the psycho-acoustic gain shaping unit 114 receiving signals from and transmitting signals to the spectral DEC unit 102. The spectral DEC unit 102 comprises a Fast-Fourier transformation (FFT) unit 115 which receives the signal x[k] and transforms it into a spectral signal X(ω). The signal X(ω) is supplied to the psycho-acoustic gain shaping unit 114 and to a multiplier 116, which also receives a signal G(ω) representing spectral gain factors from the psycho-acoustic gain shaping unit 114. The multiplier 116 provides a spectral signal Y(ω) which is fed into an inverse Fast-Fourier transformation (IFFT) unit 117 and transformed into the signal y[k].

In one embodiment, the application of linear predictive coding (LPC) in the frequency domain is very favourable. As linear predictive coding (LPC) analysis is performed in the frequency domain the equalizer gain factors are estimated in the spectral range and are dependent on the current noise signal only. Here not only can the voice signal be computationally removed from the microphone signal, but also psycho-acoustic properties can be used in a relatively simple manner when calculating the gain factors. The transition from the time to the spectral range can take place either using a Fast Fourier Transformation (FFT) or a warped Fast Fourier Transformation (WFFT), or with a psycho-acoustically motivated filter bank with frequency bands of different widths (e.g., based on a bark scale).

Linear predictive coding is one of a family of analysis/re-synthesis techniques developed for speech encoding, transmission and reception and is still the most widely used technique for computer-generated speech. LPC attempts to estimate the spectrum of a sound in terms of the filter coefficients that would be needed to synthesise that sound when applied to an excitation source, which may be either noise (for unvoiced sounds), or a band-limited pulse wave in which all the harmonics are of equal amplitude. In a synthesis application, both the excitation waveform and the filter coefficients can be modified (in both the frequency and time domains) to create related variants of the original sound. (see Richard Dobson, A Dictionary of Electronic and Computer Music Technology, Oxford University Press, 1992)

When calculating the gain factors (G(∝)), the spectrum of the desired signal (X(ω)) can be used along with the estimated noise spectrum (N(ω)). In doing so, masking effects, for example, can be considered in addition to a detailed reproduction of the ear characteristics, whose mean value as manifested in an A-filter as used in time range solutions. The differentiating frequency resolution property of the human ear can be considered as early as during the transition from the time range to the spectral range—e.g., using a psycho-acoustic motivated filter bank or a WFFT, but this is not mandatory. A conventional FFT may also be used and the spectral lines then processed in groups as well as a variable tap profile for subband adaptive filtering (see, for example, F. Weiss, R. W. Stunt, On Adaptive Filtering in Oversampled Subbands, 1998, ISBN 3/8265/45/66/4).

In the system of FIG. 1, the music signal is conventionally extracted from the microphone signal using an adaptive filter operated in the time domain. The consequence of this approach is that only the superimposition of the necessary noise signal with the noisy speech signal remains in ideal cases. The noise effects of the speech signal are then suppressed in the spectral domain using a voice activity detection mechanism so that in ideal cases the noise signal remains (as required). The spectral shape of this noise signal is not affected to any major extent, a factor that is necessary in order to then, together with the source signal, be able to compute the equalizer filter. Psycho-acoustic characteristics can be used in a relatively simple manner—this aspect will be considered in greater detail later.

A further distinguishing factor of a spectral DEC system as shown in FIG. 1, is that the FIR filtering—i.e., the weighting of the desired signal (music signal) with the equalizer filter using the fast convolution, is also carried out directly in the spectral domain. Depending on the length of the FIR filter or the FFT (Fast Fourier Transformation), this solution can considerably shorten the computing time. A general rule of thumb is that FIR filters with a length of about 30 taps can be implemented with greater computing efficiency using the fast convolution than in the time domain. However, the fast convolution can significantly increase memory requirements under certain circumstances. This is not just a problem associated with the fast convolution; it also occurs with every form of signal processing in the spectral domain. Either an FFT or warped FFT (WFFT) or a psycho-acoustic-motivated filter bank with frequency bands of different widths can be used for the transition from the time to the spectral domain. The frequency resolution characteristics of the human ear may already be considered here. However, as already mentioned, a conventional FFT also can be used. In this case, the spectral lines must be subsequently processed as groups in the spectral domain in order to account for the frequency resolution characteristics of the human ear.

An alternative implementation of a DEC system in the spectral domain is illustrated in FIG. 2. It differs from the one shown in FIG. 1 in that the adaptive filter is also implemented in the spectral domain. There are different ways of implementing an adaptive filter in the spectral domain, but only the best-known solution, the so-called overlap save FDAF (Frequency Domain Adaptive Filter), is referred to in this alternative system illustrated in FIG. 2.

In the system of FIG. 2, a signal source 201 supplies a desired signal, for example a music signal x[k] from a CD player, radio, cassette player or the like, to a spectral dynamic equalization control (DEC) unit 202 operated in the frequency domain and providing an equalized signal OUT[k] on a line 207 to a loudspeaker 203. The loudspeaker 203 generates an acoustic signal that is transferred to a microphone 204 wherein the transferal can be described by a transfer function H(z). The signal from the microphone 204 is supplied via a spectral voice suppression unit 209 and a psycho-acoustic gain shaping unit 214 (both operated in the frequency domain) to the DEC unit 202.

The voice suppression unit 209 comprises a Fast-Fourier transform (FFT) unit 210 for transforming signals from the time domain into the frequency domain. In a mean calculation unit 211 the signals in the frequency domain from the FFT unit 210 are averaged and supplied to a voice activity decoder (VAD) 212 for detecting spectral voice signal components in the signals in the frequency domain from the FFT unit 210.

The signals from the VAD 212 are supplied to the psycho-acoustic gain shaping unit 214 receiving signals from and transmitting signals to the spectral DEC unit 202. The spectral DEC unit 202 comprises a Fast-Fourier transformation (FFT) unit 215 which receives the signal x[k] and transforms it into a spectral signal X(ω). The signal X(ω) is supplied to the psycho-acoustic gain shaping unit 214 and to a multiplier 216 which also receives a signal G(ω) representing spectral gain factors from the psycho-acoustic gain shaping unit 214. The multiplier 216 generates a spectral signal OUT(ω) the signal OUT(ω) is fed into an inverse Fast-Fourier transformation (IFFT) unit 217 and transformed to provide the signal OUT[k].

In FIG. 2, instead of the adaptive filter for estimating the noise in the microphone signal operated in the time domain, an adaptive filter 206 operated in the frequency domain receives the microphone signal representing the total sound level in the environment, wherein the total sound level comprises both the sound output from the loudspeaker 203 and the ambient noise within the environment (LRM system). The signal X(ω) is used as a reference signal for the adaptive filter 206. The signal output by the adaptive filter 206 is transferred to an inverse Fast-Fourier transformation (IFFT) unit 219 and transformed into the signal y[k]. A difference unit 232 computes the difference between the signal on the line 230 and the output signal from the microphone to generate an error signal on a line 234 indicative of the difference.

The overlap save FDAF as shown in FIG. 3 can be implemented both with and without the so-called constraint—i.e., the rectangular windowing of the adaptive filter coefficients in the time domain. Along with considerably less computing time (because 2 FFTs can be omitted), the only other difference is particularly evident in the convergence speed of the two alternative methods. The method with the constraint is adapted significantly faster than the method that does without restrictions on filter coefficients to possibly speed up computing time.

The overlap-safe FDAF comprises an input block processing unit 301 receiving a signal x[n] as an input signal. The input block signal includes an old block and a new block wherein the new block is attached to the old block. The signal output from the input block signal unit 301 is supplied to a Fast Fourier Transformation (FFT) unit 302 that provides a signal X(k) in the frequency domain which corresponds to the input signal x[n] in the time domain. The signal X(k) is supplied to a multiplier 303 and to a complex conjugate processing unit 304, the output signal X′(k) of which is supplied to a multiplier 305. The multiplier 305 also receives a signal E(k) from a FFT-unit 306 and supplies its output signal to a multiplier 307 for multiplying with a constant 2μ. An adder 308 is connected downstream of the multiplier 307, and the adder 308 also receives its own output signal W(k) delayed by a delay unit 309 as a delayed signal W(k−1).

The delayed output signal W(k−1) is supplied to a constraint unit 310 that provides one of the signals supplied to the multiplier 303. The output signal Y(k) is transformed by an inverse Fast Fourier Transformation (FFT) unit 311 into an output signal in the time domain wherein only the second half of the output block signal is used for forming an output signal y[n]. Selecting of the second half of the output block signal is performed in an output block signal unit 312. The output signal y[n] and the desired response signal d[n] are supplied to a subtractor 313 that generates an error signal e[n] therefrom. In a processing unit 314 connected downstream to the subtractor 313 and receiving the error signal e[n], a block containing only zero bits is attached to the error signal forming an error block signal. The error block signal is supplied to the FFT unit 306. The optional constraint unit 310 comprises an inverse Fast Fourier Transformation (IFFT) unit 315, a block delete unit 316 for deleting the last data block, a block attach unit 317 for attaching a block containing only zero bits, and a Fast Fourier Transformation (FFT) unit 318 connected in series between the delay unit 309 and the multiplier 303.

One advantage of an adaptive filter implemented in the spectral domain is that it is markedly more stable than its time domain counterpart—i.e., it cannot be so quickly disturbed, for example, by strong noise signals as only the spectral components of the adaptive filter are affected by noise at the exact location of the noise. If the noise relates, for example, to a speech signal, it usually will not have a broadband spectrum, but mainly includes different narrowband signals, so-called formants. Although the duration of the formants is short, they contain enough energy to usually disturb the adaptive filter at the point they occur—i.e., without any countermeasure.

Spectral domains beside the formants are either not disturbed at all or not so strongly disturbed that the adaptive filter is consistently disturbed—in other words, the approximated model retains its correct functionality for the most part. This circumstance is also the reason why an adaptive adaptation step size might not even be needed in certain cases for adaptive filters implemented in the spectral domain (the adaptive step size is required for a time domain filter). However, if an adaptive adaptation step size is involved and therefore the adaptation for the spectral domains affected by a large noise signal for the duration of the disturbance is omitted, the already approximated model can be broadband protected against destruction by overly large noise factors. However, there may be the risk of complete adaptation blockage—for example, in the event of sudden changes in the room impulse response, which demands suitable countermeasures.

One solution to this problem is the normal or standardized coherence function between the source signal (x[n]) and the microphone signal (d[n]) or between the output signal of the adaptive filter core (y[n]) and the microphone signal (d[n]) in the frequency range. Better results can be obtained, however, by using a second alternative method: the so-called double-talk detector. Ideally, the optimum adaptation step size is obtained from the relationship between the spectral power densities of the residual echo (b[n]) and the error signal (e[n]). In practical terms, however, direct computation is not possible because the required residual echo signal is inaccessible. This signal can nonetheless be estimated. The power spectral density of the residual echo signal can be estimated in a simple manner by filtering the power spectral density of the input signal with the power spectral density of the current echo path model (W(k)). The following equation applies in this respect:

u(n,k)=Φbb(n,k)Φee(n,k)=W(n,k)2*Φxx(n,k)Φee(n,k)(EQ.1)

Wherein equation 1 is a computation of the adaptive adaptation step size in the spectral domain.

If equation 1 is used to compute the adaptive adaptation step size, it is now possible not only to solve the problem of disturbance of the approximated model due to strong noise signals, but also to remedy the problem of sudden changes in the room impulse response. This can be easily explained using equation 1 as shown above: although the error signal rises in situations with strong noise, the residual echo does not, and consequently the adaptation step size falls. In contrast, in the case of sudden changes of the loudspeaker-room-microphone (LRM) system, both the residual echo and the error signal increase, which means the adaptation step size has a large value, which is necessary for new adaptations.

Whenever music is presented in a noisy environment, the noise is considered a disturbance. In the past measures against noise have been taken. Before DVC a user simply turned up the volume on the sound system as appropriate to counteract the noise. The DVC system then automatically took over doing this task. It was discovered at the time of using the DEC system in the time domain that not only the loudness has to be adapted according to the noise, but also that its spectral coloring through an appropriate equalizing filter should be considered so that most of the original audio characteristic of the music signal is retained. The equalizing filter was automatically set using different types of analysis in the time domain in such a way that it roughly follows the power spectral density (PSD) of the noise signal. Accordingly a system is wanted that is not only able to increase the loudness as required, but also to appropriately counteract the spectral distortion of the music signal as well. A model of the equalizer filter that is based solely on the PSD of the noise signal is not considered suitable as it would ignore the currently available music signal. A better system would appear to be one that produces the equalizer filter using both the noise and music signals.

For a better understanding, the following example is considered: a 1-kHz tone reduced by 12 dB is to be used as a noise signal, and white noise reduced by 20 dB is to serve as the music signal. Neither increasing the loudness of the music signal alone, nor just equalizing with the PSD of the noise signal nor a combination of these two approaches would provide an acceptable solution. The reason for this is to be found in the psycho-acoustics and is referred to as ‘masking’ (see, for example, Zwicker E., Fastl H., 1998, Psychoacoustics, Facts and Models, Springer Series in Information Sciences, Springer-Verlag, second edition). This term means that human hearing is so affected by narrowband noise signals that not only the band identical to the noise signal, but also adjacent frequency bands are influenced. Consequently, in the above example an equalizer filter with a bandpass characteristic of low quality and a center frequency equal 1 kHz is required. However, the DEC system in the time domain would not provide us with such an equalizing function.

To find an adequate equalizer filter, the masking threshold value of the noise signal has to be evaluated, first. Further, the music signal also has a masking threshold value, which has to be considered in the computation of the equalizer filter because spectral domains in which the music signal masks the noise signal—i.e., renders inaudible, are not permitted to be increased if the original audio characteristic of the music signal is to be retained. More important is the difference between the two masking thresholds as only the spectral domains in which the noise signal masks the music signal are allowed to appear in the equalizer filter.

FIG. 4 shows the characteristics of masking thresholds compared to the absolute background threshold. One of the curves depicted in the graph is the masking threshold of a noise signal typically experienced in an automobile, while the other curve shows the masking threshold of a typical music signal. The characteristic of the gain function is computed from the difference between the masking thresholds of the noise and music signals—i.e., the characteristic of the equalizer filter, which is shown below for this example.

FIG. 5 shows a representation of a typical equalizer filter in an automobile. The disadvantage of this method is that the gain characteristic rises infinitely as soon as the music signal pauses or becomes very quiet. This can be solved by limiting the maximum gain. In addition, increasing the unavailable music signal would admittedly cause an increase in (quantization) noise, but only up to the masking threshold of the noise signal, which would not be noticed in ideal cases since, on the one hand, it would be masked by the noise signal and, on the other hand, would exhibit its spectral form.

A number of masking models are known, two of which have established themselves in practical use. One is the model based on the MPEG format (see, for example, ISO/IEC JTC1/SC29/WG11 MPEG, 1993, IS11172-3, Information technology—Coding of moving pictures and associated audio for digital storage at up to about 1.5 Mbit/s—Part 3: Audio, ISO/IEC), which is also applied in the familiar MP3 music format, while the other is the so-called Johnston model (see, for example, Johnston J. D., 1988, Estimation of Perceptual Entropy Using Noise Masking Criteria, Proc. ICASSP'88, pp. 2524-2527, or Johnston J. D., 1988, Transform Coding of Audio Signals Using Perceptual Noise Criteria. IEEE Journal on Selected Areas in Communications 6, pp. 314.323), which is favored for the present implementation because it is easier to scale. In contrast, the MPEG-based model uses fixed tables available only for a limited number of sampling frequencies. It is sophisticated to modify these tables, for example, to obtain any sampling frequency required. This is a factor that makes the Johnston model preferable.

In practice, the equalizer filter is usually specified for one channel, but is nonetheless used for multiple channels. Therefore, an equalizer filter is needed—both in the frequency and time domains—with different source signals. Filtering in the time domain can be accomplished using an FIR filter computation, or in the spectral domain using fast convolution. Since the equalizer filter is a long filter (due to its broadband nature), a direct implementation in the frequency domain requires too much memory, and a direct implementation in the time domain takes up too much computation time. The former problem can be regarded as the more pressing because equalizer filtering takes place in the spectral domain and therefore requires a large amount of memory.

Thus the long FIR equalizer filter needs to be replaced by a less complex one. Referring to the LPC analysis in the time domain by iteratively computing the LPC coefficients, for example, using the GAL (Gradient Adaptive Lattice) algorithm, and then inserting the computed coefficients in a pure predictor filter (i.e., IIR filter). In this way, the length of the required filters can be reduced without giving rise to grave problems. Below is a schematic representation of how such a computation of the IIR filter coefficients using the GAL algorithm may appear.

FIG. 6 is a representation of the iterative computation of the IIR filter coefficients using the GAL algorithm. In FIG. 6, a signal x[k], essentially a white noise signal from a noise source 601 is supplied to an equalizing filter 602 in the time domain. The filter coefficients for the equalizing filter 602 are derived by a psycho-acoustically masking model. The signal output by the equalizing filter 602 is supplied to a GAL-algorithm unit 603. The GAL-algorithm unit 603 provides reflection coefficients that are transformed into direct form coefficients in a conversion unit 604. These direct form coefficients are used as filter coefficients in an IIR-filter 605 which is connected downstream of a desired-signal source 606 and providing an output y[n] on a line 610.

FIG. 7 illustrates the comparison of the original equalizer filter with the approximated equalizer filter. FIG. 7 shows the gain of an equalizer filter typically found in an automobile. The gain curve referred to as the original represents the gain directly computed from the psychoacoustic model, while the dotted curve (referred to as the approximation) represents the original gain's reproduction using an IIR filter. The original gain curve was generated using a 512-tap FIR filter, the approximation using a 50-tap IIR filter. In view of the fact that the factor is reduced to a tenth, the general shape of the approximated curve matches the original quite well. This proves that the method can be used when there is a shortage of computing time and memory.

In FIG. 8, a DEC-system 801, which is a similar system to those systems shown in FIGS. 1 and 2, is supplied with a desired signal x[n] via a sampling rate decimeter 802. The output signal of the DEC-system 801 is supplied to a loudspeaker 803 via a sampling rate interpolator 805. The loudspeaker 803 generates an acoustic signal from an output signal OUT[n] and the acoustic signal is transferred to a microphone 804 wherein the transferal can be described by a transfer function H[z]. The signal d[n] from the microphone 804 is supplied via a sampling rate decimeter 806 to the DEC-system 801. The signal x[n] is bypassed in a bypass path comprising highpass filter 807, a delay unit 808, and a controllable gain unit 809. The controllable gain unit 809 is controlled by the DEC-system 801 which internally is controlled by a volume mechanism.

One can see that although an implementation of the DEC system in the spectral domain is more complex than one in the time domain, the processing benefits from greater flexibility and possibilities, which in the final analysis results in higher quality of the system.

The above-mentioned systems may be implemented in microprocessors, signal processors, microcontrollers, computing devices etc. The individual system components are in this case hardware components of the microprocessors, signal processors, microcontrollers, computing devices, etc. which are correspondingly implemented by software.

In the following some aspects of the invention are summarized. One example of the invention may be:

  • (a) An audio enhancement system for compensating for ambient noise in a listening environment, comprising
    • audio means for producing an electrical sound signal in the time domain and generating a sound output from said electrical sound signal;
    • said audio means comprising a time-to-frequency transformation means responsive to said electrical sound signal in the time domain for generating an electrical sound signal in the frequency domain;
    • said audio means further comprising frequency-to-time transformation means responsive to said electrical sound signal in the frequency domain for generating a re-transformed electrical sound signal in the time domain for generating the sound output;
    • sensor means for obtaining a total sound signal representative of the total sound level in said environment, wherein said total sound level comprises both said sound output from said audio means and the ambient noise within said environment;
    • extraction means responsive to said total sound signal and to said electrical sound signal for extracting an ambient noise signal representative of the ambient noise in said environment from said total sound signal; and
    • control means responsive to said ambient noise signal and said electrical sound signal in the frequency domain for equalizing and adjusting in the frequency domain the sound output of said audio means to compensate for said ambient noise level.
  • (b) The audio enhancement system of point (a), wherein the control means comprises a gain shaping unit for performing an equalizing function in the frequency domain.
  • (c) The audio enhancement system of point (a) or (b), wherein the control means comprises a voice activity detector.
  • (d) The audio enhancement system of point (a), (b) or (c), wherein the control means comprises a mean value calculation unit.
  • (e) The audio enhancement system of one of the points (a) to (d), wherein said control means comprises a time-to-frequency transformation unit connected to said sensor means for transforming said total sound signal in the time domain into a total sound signal in the frequency domain.
  • (f) The audio enhancement system of point (e), wherein said extraction means is a noise estimation unit operated in the frequency domain; said noise estimation unit being included in the control means.
  • (g) The audio enhancement system of point (f), wherein said noise estimation unit is a frequency domain adaptive filter.
  • (h) The audio enhancement system of point (g), wherein said frequency domain adaptive filter is an overlap-save frequency domain adaptive filter.
  • (i) The audio enhancement system of point (e), wherein said extraction means is a noise estimation unit operated in the time domain; said noise estimation unit being connected between the sensor means and the control means.
  • (j) The audio enhancement system of point (i), wherein said noise estimation unit operated in the time domain comprises at least one adaptive filter.
  • (k) The audio enhancement system of one of the points (a) to (j), wherein said control means performs a linear predictive coding in the frequency domain.
  • (l) The audio enhancement system of point (b), wherein said gain shaping unit performs a psycho-acoustically motivated gain shaping function.
  • (m) The audio enhancement system of point (l), wherein the gain shaping unit has a frequency scale according to psycho-acoustic frequency groups.
  • (n) The audio enhancement system of point (l), wherein said gain shaping unit has a warped frequency scale.
  • (o) The audio enhancement system of point (l), wherein said gain shaping unit performs psycho-acoustic masking.
  • (p) The audio enhancement system of point (o), wherein said psycho-acoustic masking is performed in the MPEG format.
  • (q) The audio enhancement system of point (o), wherein said psycho-acoustic masking is performed according to the Johnston model.
  • (r) The audio enhancement system of one of the points (a) to (q), wherein the audio means comprises a bypass path for higher frequencies; said bypass path having a controllable gain controlled by the control means.
  • (s) The audio enhancement system of point (r), wherein said bypass path comprises a delay unit.
  • (t) The audio enhancement system of point (r) or (s), wherein the control means is operated in an undersampled mode.

Another example of the invention may be:

  • (A) A method for enhancing the sound signal produced by an audio system in a listening environment by compensating for ambient sound in said listening environment, comprising the steps of
    • producing an audio sound in the time domain from an electrical sound signal in the time domain; said electrical sound signal in the time domain being transformed into electrical sound signal in the frequency domain and said electrical sound signal in the frequency domain being retransformed into audio sound in the time domain;
    • measuring the total sound level in said environment and generating a signal representative thereof,
    • processing the audio sound signal and the total sound signal using an algorithm to extract a signal representing the ambient sound level within said environment;
    • performing an equalization in the frequency domain and adjusting the output from said audio sound signal to compensate for said ambient noise level.
  • (B) The method of point (A), wherein the equalization step comprises a gain shaping function in the frequency domain.
  • (C) The method of point (A) or (B), wherein the equalization step comprises a voice activity detection step.
  • (D) The method of point (A), (B) or (C), wherein the equalization step comprises a mean value calculation step.
  • (E) The method of one of the points (A) to (D), wherein said equalization step comprises a time-to-frequency transformation step for transforming said total sound signal in the time domain into a total sound signal in the frequency domain.
  • (F) The method of point (E), wherein said extraction step is a noise estimation step in the frequency domain.
  • (G) The method of point (F), wherein said noise estimation step is an adaptive filtering step in the frequency domain.
  • (H) The method of point (G), wherein said adaptive filtering step is an overlap-save adaptive filtering step in the frequency domain.
  • (I) The method of point (E), wherein said extraction step is a noise estimation step in the time domain.
  • (J) The method of point (I), wherein said noise estimation step in the time domain comprises at least one adaptive filtering step.
  • (K) The method of one of the points (A) to (J), wherein said equalization step comprises linear predictive coding in the frequency domain.
  • (L) The method of point (B), wherein said gain shaping step is a psycho-acoustically motivated gain shaping step.
  • (M) The method of point (L), wherein the gain shaping step comprises a frequency scale according to psycho-acoustic frequency groups.
  • (N) The method of point (L), wherein said gain shaping step comprises a warped frequency scale.
  • (O) The method of point (L), wherein said gain shaping step performs psycho-acoustic masking.
  • (P) The method of point (O), wherein said psycho-acoustic masking is performed in the MPEG format.
  • (Q) The method of point (O), wherein said psycho-acoustic masking is performed according to the Johnston model.
  • (R) The method of one of the points (A) to (Q), wherein the audio means comprises a bypass path for higher frequencies; said bypass path having a controllable gain controlled by the control means.
  • (S) The audio enhancement system of point (R), wherein said bypass step comprises a delay step.
  • (T) The audio enhancement system of point (R) or (S), wherein the equalization step operates in an undersampled mode.

Although various exemplary embodiments of the invention have been disclosed, it will be apparent to those skilled in the art that various changes and modifications can be made which will achieve some of the advantages of the invention without departing from the spirit and scope of the invention. It will be obvious to those reasonably skilled in the art that other components performing the same functions may be suitably substituted. Further, the methods of the invention may be achieved in either all software implementations, using the appropriate processor instructions, or in hybrid implementations that utilize a combination of hardware logic and software logic to achieve the same results. Such modifications to the inventive concept are intended to be covered by the appended claims.