Audio enhancement system and method

BACKGROUND OF THE INVENTION

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

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 lessened. 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 audio system's volume will be too high, requiring the user to reduce it. To overcome this, US Patents 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 gain chase situation could 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 range - i.e., in a broadband manner - and is implemented by an averaging unit with permanently con-FIGs. 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 VAD's output signal. 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 which is 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 fully ideal and could 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 completely avoiding gain chase is to upward-limit the control signal in accordance with the existing power of the desired signal which 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 lessen the risk of gain chase and additionally to reduce the implementation work.

It is an object of this invention to create an improved system and method which automatically compensate for the noise level in a listening environment.

SUMMARY OF THE INVENTION

The present invention comprises a dynamic equalizer control (DEC) system and method which digitally processes electrical audio signals to compensate for the intrusion of acoustic noise into a listening area. Such a system could be used with any audio system and any listening environment. For the 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 could be used in other listening environments, such as a room, as well.

The DEC system in accordance with the present invention comprises the following elements: audio means for producing an electrical sound signal and generating a sound output from said electrical sound signal; 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; control means responsive to said ambient noise signal for performing a linear predictive coding (LPC) analysis and generating a control signal based on the outcome of said analysis; and equalizer means responsive to said control signal for adjusting the sound output of said audio means to compensate for said ambient noise level.

The DEC system in accordance with the present invention continuously 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 consists of 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 disclosed by the present invention includes hardware and appropriate software to shape the spectrum of the music and noise signal to mimic human hearing.

One major aspect of the present invention is the application of linear predictive coding (LPF) in a sound enhancement system. Linear predictive coding is one of a family of analysis/resynthesis 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 which 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)

The present 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 according to the invention therefore controls the desired signal (music signal) according to the spectral distribution of the noise signal, and in doing so psychoacoustic aspects may also be considered when configuring the equalization.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention can be better understood with reference to the following drawings and description. The components in the FIGs. are not necessarily to scale, instead emphasis being placed upon illustrating the principles of the invention. Moreover, in the FIGs., like reference numerals designate corresponding parts. In the drawings: <dl id="dl0001"><dt>FIG. 1</dt><dd>is a block diagram of a system for broadband removing of a desired signal from a sensor signal;</dd><dt>FIG. 2</dt><dd>is a block diagram of a DEC system operating in the time domain;</dd><dt>FIG. 3</dt><dd>is a block diagram of a 4th-order Adaptive Lattice Predictor (ALP) filter;</dd><dt>FIG. 4</dt><dd>is a block diagram of a predictor filter designed as a lattice ladder all-pole filter;</dd><dt>FIG. 5</dt><dd>is a block diagram of a DEC system comprising a combination of a beamformer and an adaptive filter;</dd><dt>FIG. 6</dt><dd>is a block diagram of a DEC system comprising a negative beamformer and an adaptive filter; and</dd><dt>FIG. 7</dt><dd>is a block diagram of a DEC system comprising time and spectral processing.</dd></dl>

DETAILED DESCRIPTION

Analysis shows that the main reason why adaptive filters fail in broadband applications is to be found in the wide dynamic range of the input signals. By limiting the system to a narrow frequency band, the dynamics of the band-limited signal are correspondingly reduced so that the adaptive filter can work better. If the broadband input signals are divided into frequency bands in such a way that the band-limited signals exhibit dynamic characteristics with which an adaptive filter still functions without errors, the music signal can be removed from the sensor signal in a broadband manner. Synergy effects can in turn be used to improve efficiency as a consequence. For example, the divided desired signals can be used for a multipath loudspeaker without having to spectrally split the signals yet again. This would also yield the additional advantage that the corresponding adaptive filter could also explicitly reproduce the appropriate frequency response of the corresponding loudspeaker.

As already mentioned, the strongest noise signals by far in vehicles generally occur in the low frequency range, which means limiting to a frequency range, which however is much larger, without having to worry about the loss of effects. Since the sensor signal freed from the desired signal which is obtained from the adaptive filter is to be recompiled, the filters for splitting the signal into different bands must be perfect reconstruction or quasi-perfect reconstruction filters. An effective method of implementing a quasi-perfect reconstruction filter for realization of the band splitting (i.e., crossover or X-over) filters is to use so-called coupled all-pass filters.

FIG. 1 illustrates a system for broadband removing of a desired signal from a sensor signal. A signal source 101 supplies a desired signal, e. g., a music signal from CD player, radio, cassette player or the like, to a processing unit 102 as an equalizing filter for equalizing the music signal according to a control signal supplied to the processing unit 102. A first X-over network 103 is connected downstream of the processing unit 102 splitting the output signal from the processing unit 102 into a first higher frequency signal and into a first lower frequency signal. The first higher frequency signal is fed into second X-over network 104 and, accordingly, split into a second higher frequency signal and into a second lower frequency signal. The first lower frequency signal, the second lower frequency signal, and the second higher frequency signal are supplied to active loudspeakers 108, 109, 110 respectively via sample rate converters 105, 106, 107 respectively for changing the respective sample rates. The loudspeakers are located in a room together with a microphone 111 as sensor for obtaining a total sound signal representative of the total sound level in said room (environment). The loudspeaker, room, and microphone form a so-called loudspeaker-room-microphone (LRM) system 112 which comprises certain transfer fuctions H_Sub, H_Bass, H_mid between the respective loudspeakers 108, 109, 100 and the microphone 111. The signal representing the total sound level comprises both said sound output from said audio means and the ambient noise n[k] generated said room, i. e., the LRM system 112.

The microphone 111 is connected to extraction means responsive to the total sound signal. The extraction means comprises a third X-over network 112 receiving the total sound signal and splitting it into a third lower frequency signal and a third higher frequency signal wherein the third higher frequency signal is splitted into a fourth lower frequency signal and a fourth higher frequency signal by means of a fourth X-over network 113. The third lower frequency signal, fourth lower frequency signal, and fourth higher frequency signal are supplied via sample rate converters 114, 115, 116 respectively to three adaptive filters 117, 118, 119 which also receive the signals output by the sample rate converters 105, 106, 107. Each of the adaptive filters 117, 118, 119 comprises a filter core 120, 121, 122 receiving the signals output by the sample rate converters 105, 106, 107 respectively and control signals from control units 123, 124, 125 respectively. The control units 123, 124, 125 are supplied with the signal output by the sample rate converters 105, 106, 107 and signals representing the difference of the signals output by the the sample rate converters 114, 115, 116 and the signals output by the filter cores 120, 121, 122 respectively. Said signals representing the difference are fed into a summing unit 126 126 generating an approximated noise signal n^-[k] therefrom; said approximated noise signal n^-[k] being supplied to the processing unit 102 as its control signal.

Assuming that the system illustrated in FIG. 1 is able to remove the desired signal from the sensor signal (microphone signal) so efficiently that only a small portion of the signal is left so that no more gain chase can occur, the question still remains as to how an equalizing filter from the sensor signal (without the desired signal) can be obtained. An effective means of doing this in the time range is according to the invention to compute the Linear Predictive Coding (LPC) coefficients using LPC analysis, and to input the coefficients into a predictor filter with which the music signal is weighted.

It is known from other applications - e.g., acoustic echo cancellation (AEC) - that echo coefficients are calculated from the voice signal in those cases using LPC analysis, and that the coefficients are then fed into a predictor error filter for filtering the microphone signal. The voice signal recorded by the sensing microphone is spectrally smoothed so that the AEC system can work more efficiently. Said AEC application therefore refers to a so-called pre-whitening filter whose effects can be described as follows: the LPC analysis supplies coefficients that are used in a corresponding predictor filter that calls a transfer function corresponding to the power spectral density (PSD) of the signal (i.e., voice signal) examined by the analysis. When used in a predictor error filter, its transfer function does not follow the PSD of the analysis signal, but rather the inverse pattern. For this reason, the voice component in the microphone signal on which the LPC analysis was based is spectrally smoothed after the predictor error filter - i.e., it is whitened (thus giving rise to the term "pre-whitening filter").

The analysis/synthesis method known as linear predictive coding (LPC) is an efficient and effective mean, for example, to achieve synthetic speech and speech signal communication The efficiency of the method is due to the speed of the analysis algorithm and to the low bandwidth required for the encoded signals. The effectiveness is related to the intelligibility of the decoded vocal signal. The LPC implements a type of vocoder which is an analysis/synthesis scheme where the spectrum of a source signal is weighted by the spectral components of the target signal that is being analyzed. The phase vocoder is, for example, a special kind of vocoder where amplitude and phase information of the analysis channels is retained and can be used as coefficients for an error filter.

In the standard formulation of LPC, the source signals are either a white noise or a pulse train, thus resembling voiced or unvoiced excitations of the vocal tract, respectively. The basic assumption behind LPC is the correlation between the n-th sample and the P previous samples of the target signal. Namely, the n-th signal sample is represented as a linear combination of the previous P samples, plus a residual representing the prediction error. The analysis problem is equivalent to the identification of the coefficients a of an allpole filter. If minimizing the error in a mean square sense, the problem translates into a set of P equations which, in the z domain (frequency domain), reduce to<maths id="math0001" num=""><math display="block"><mrow><mtext>E(z) = A(z)X(z)</mtext></mrow></math><img file="EP1580882A1_D0001.tif" /></maths> where A(z) is the polynomial with coefficients a and E(z) is the prediction error. In the case of voice signal analysis, the filter 1/A(z) is called the allpole formant filter because, if the proper order P is chosen, its magnitude frequency response follows the envelope of the signal spectrum, with its broad resonances called formants. The filter A(z) is called the inverse formant filter because it extracts from the voice signal a residual resembling the vocal tract excitation. A(z) is also called a whitening filter because it produces a residual having a flat spectrum.

However, there are two kinds of residuals, both having a flat spectrum: the pulse train and the white noise, the first being the idealized vocal-fold excitation for voiced speech, the second being the idealized excitation for unvoiced speech. In reality, the residual is neither one of the two idealized excitations. At the resynthesis stage the choice is either to use an encoded residual or to choose one of the two idealized excitations according to a voiced/unvoiced decision made by the analysis stage.

When the target signal is periodic (voiced speech), a pitch detector can be added to the analysis stage, so that the resynthesis can be driven by periodic replicas of a basic pulse, with the correct inter-pulse period.

Several techniques are available for pitch detection, either using the residual or the target signal. Although not particularly efficient, one possibility is to do a Fourier analysis of the residual and estimate the fundamental frequency by the techniques of section. For example, the information extracted by the analysis stage are the prediction coefficients a₁,...a_p; the residual e; pitch of the excitation residual; voiced/unvoiced information; and signal energy (RMS amplitude). These parameters, possibly modified, are used in the resynthesis stage, for example, by means of the well-known Levinson-Durbin recursion, which provides the reflection coefficients of the lattice realization of the filter 1/A(z).

FIG. 2 is a block diagram of a DEC system operating in the time domain. Such DEC system may be implemented in the processor unit 102 of FIG. 1. The approximated noise signal n^-[k] representing the microphone (sensor) signal cleansed of the music signal is submitted to an LPC analysis unit 201, and the LPC coefficients (echo coefficients) obtained are input into a predictor filter 202, with which a music signal x[k] from a signal source like the signal source 101 of FIG. 1 is filtered. In this way, the music signal x[k] is weighted or equalized in accordance with the existing power spectral density of the analysis signal. To take account of human hearing, there are two major factors to consider. On the one hand, the human ear is sensitive both to frequency and loudness: the maximum sensitivity is about 1-2 kHz; it deteriorates both toward higher as well as lower frequencies.

This relationship is considered in FIG. 2 using a psycho-acoustic filter 203 (for example, an A-weighting filter), which approximately matches the mean value of the human ear's audio characteristics. On the other hand, the human ear has a frequency resolution that increases with decreasing frequency. This characteristic is to be considered by the LPC analysis. Its accuracy would therefore also increase with decreasing frequency. A corresponding non-linear LPC analysis function in terms of frequency can be realized using the so-called warped LPC (WLPC) analysis where the predictor filter is also configured as a warped predictor filter. Using such system, a weighted output signal (y[k]) is obtained that varies according to the current noise signal, regardless of whether the system has a psychoacoustic filter or not, or whether it uses conventional LPC analysis or WLPC analysis. Different ways to implement conventional or warped LPC are, for example, disclosed in Aki Härmä, "Linear predictive coding with modified filter structures"; Aki Härmä, "Evaluation of a warped linear predictive coding scheme"; Aki Härmä, Unto K. Laine," A comparison of warped and conventional linear predictive coding" (available on the internet under www.acoustics.hut.fi) which are herein included by reference.

One problem that has been completely ignored so far is that the voice signal, which is still contained in the remaining signal n^-[k], has the effect in the DEC model described above that the equalizing varies according to the voice signal. Although this may not cause any inconvenience for a person talking, for whom the equalizing is altered according to the voice signal, as this represents a type of masking for the speaker, it can be problematic for all other persons in the same environment. The equalizing caused by the voice has an annoying effect - for example, because the addressed persons want to clearly understand the speaker's voice and do not want the voice to be dominated by the music signal. One solution is to set the update time of the LPC analysis in such a way that the LPC coefficients adapt themselves according to the background noise signal (which changes slowly in comparison to the voice signal) and do not or only slightly respond to the voice signal.

An efficient computational method for the LPC coefficients is provided by the so-called GAL (Gradient Adaptive Lattice) algorithm, with which any number of LPC coefficients can be recursively computed without using the otherwise very complexe Levinson-Durbin algorithm. The filter structure with which the GAL algorithm can be realized is known by the designated ALP (Adaptive Lattice Predictor).

An applicable 4th-order adaptive lattice predictor (ALP) filter is illustrated in FIG. 3. This filter comprises three stages 310, 320, 330 of identical structure. Each stage has two branches wherein The first branch transmitts signals f₀[k], f₁[k], f₂[k], f₃[k], and the the second branch transmitts signals b₀[k], b₁[k], b₂[k], b₃[k]. Each second branch comprises a delay unit 311, 321, 331 and both branches comprise each a subtractor 312, 313; 322, 323; 332, 333. Said subtractors are used for cross-coupling of the two branches via controllable coefficient elements 314, 315; 324, 325; 334, 335 so that their coefficients K₁, K₂, K₃ are controlled by signals K₁[k], K₂[k], K₃[k]. The signals f₀[k] and b₀[k] are obtained from the signal n^-[k].

The equations with which the coefficients of the ALP filter, K₁,..., K_N-1 (which resemble the LPC coefficients) are updated are similar to the GAL algorithm. The GAL Algorithm can be described as follows:<maths id="math0002" num=""><math display="block"><mrow><msub><mrow><mtext mathvariant="italic">K</mtext></mrow><mrow><mtext mathvariant="italic">n</mtext></mrow></msub><mtext> [</mtext><mtext mathvariant="italic">k</mtext><mtext> +1] = </mtext><msub><mrow><mtext mathvariant="italic">K</mtext></mrow><mrow><mtext mathvariant="italic">n</mtext></mrow></msub><mtext> [</mtext><mtext mathvariant="italic">k</mtext><mtext>] + </mtext><mfrac><mrow><mtext mathvariant="italic">α</mtext></mrow><mrow><msub><mrow><mtext mathvariant="italic">P</mtext></mrow><mrow><mtext mathvariant="italic">n</mtext></mrow></msub><mtext>[</mtext><mtext mathvariant="italic">k</mtext><mtext>]</mtext></mrow></mfrac><mtext> * (</mtext><mtext mathvariant="italic">ƒ</mtext><msub><mrow><mtext></mtext></mrow><mrow><mtext mathvariant="italic">n-</mtext><mtext>1</mtext></mrow></msub><mtext>[</mtext><mtext mathvariant="italic">k</mtext><mtext>] *</mtext><msub><mrow><mtext mathvariant="italic">b</mtext></mrow><mrow><mtext mathvariant="italic">n</mtext></mrow></msub><mtext> [</mtext><mtext mathvariant="italic">k</mtext><mtext>] +</mtext><msub><mrow><mtext mathvariant="italic">b</mtext></mrow><mrow><mtext mathvariant="italic">n-</mtext><mtext>1</mtext></mrow></msub><mtext>[</mtext><mtext mathvariant="italic">k</mtext><mtext> -1] * </mtext><msub><mrow><mtext mathvariant="italic">ƒ</mtext></mrow><mrow><mtext mathvariant="italic">n</mtext></mrow></msub><mtext>[</mtext><mtext mathvariant="italic">k</mtext><mtext>])</mtext></mrow></math><img file="EP1580882A1_D0002.tif" /></maths><i>with</i><maths id="math0003" num=""><math display="block"><mrow><msub><mrow><mtext mathvariant="italic">P</mtext></mrow><mrow><mtext mathvariant="italic">n</mtext></mrow></msub><mtext>[</mtext><mtext mathvariant="italic">k</mtext><mtext>]=(1-</mtext><mtext mathvariant="italic">α</mtext><mtext>)*</mtext><msub><mrow><mtext mathvariant="italic">P</mtext></mrow><mrow><mtext mathvariant="italic">n</mtext></mrow></msub><mtext>[</mtext><mtext mathvariant="italic">k</mtext><mtext>-1]+</mtext><mtext mathvariant="italic">α</mtext><mtext>*(</mtext><msub><mrow><mtext mathvariant="italic">ƒ</mtext></mrow><mrow><mtext mathvariant="italic">n</mtext><mtext>-1</mtext></mrow></msub><mtext>[</mtext><mtext mathvariant="italic">k</mtext><msup><mrow><mtext>]</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><mtext> +</mtext><msub><mrow><mtext mathvariant="italic">b</mtext></mrow><mrow><mtext mathvariant="italic">n</mtext><mtext>-1</mtext></mrow></msub><mtext>[</mtext><mtext mathvariant="italic">k</mtext><msup><mrow><mtext>-1]</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><mtext>),</mtext><mtext mathvariant="italic">n</mtext><mtext> =1,...,</mtext><mtext mathvariant="italic">N</mtext><mtext>-1</mtext></mrow></math><img file="EP1580882A1_D0003.tif" /></maths>

The LPC coefficients determined using the ALP filter are input into a predictor filter, which may be designed as a lattice ladder all-pole filter as shown in FIG. 4.

The Lattice Ladder All-Pole Filter of FIG. 4 comprises three stages 410, 420, 430 of identical structure. Each stage has two branches wherein a return branch transmitts signals f₀[k], f₁[k], f₂[k], f₃[k], and a forward branch transmitts signals b₀[k], b₁[k], b₂[k], b₃[k] wherein f₃[k] = x[k] and f₀[k] = b₀[k] = y[k]. The return branches comprise each a delay unit 411, 421, 431 and an adder 413, 423, 433. Each one of the forward branches comprises a subtractor 412, 422, 432. Said subtractors and adders are used for cross-coupling of the two branches via controllable coefficient elements 414, 415; 424, 425; 434, 435 so that their coefficients K₁, K₂, K₃ are controlled by signals K₁ [k], K₂ [k], K₃ [k] .

The time range solution is not an optimum solution as it labors under two serious weaknesses. Firstly, it is doubtful that the voice signal component still in the remaining signal n^-[k] can be removed simply by using a lengthy update time. Secondly, an equalizing function that corresponds to the noise PSD characteristics either with or without A-filter weighting is not satisfactory. This can clearly be seen when considering a typical noisy situation in an automobile, for example. In this case, the power spectral density of the noise signals is somewhat pink - i.e., equalizing configured in this way would mainly raise the lower spectrum range, but omit the middle and upper ranges. As a consequence, the desired signal would be colored by a listener in that automobile and a signal would be obtained that can partially deviate strongly from the original signal, a situation that is generally undesirable.

Options for prevention of voice components in the microphone or remaining Signal are discussed in the following with reference to FIGs. 5-8.

FIG. 5 illustrates a combination of a beamformer and an adaptive filter. Beamformers are electronically steerable arrays of microphones which in combination with a phase shift network can put emphasis on a certain direction and in combination with location estimator algorithms pinpoint the location of a talker in a room. Different kinds of beamers are known, for example, from US20030053639A1. The present beamformer can be implemented either in software (beamforming filter) or hardware (for example, pressure gradient microphones, pipe slit microphones, etc.).

In in the DEC system of FIG. 5, a music signal x[k] from a music signal source 501 is fed into dynamic equalization control (DEC) unit 502 providing an equalized signal y [k] to a loudspeaker 503. From the signal y[k], the loudspeaker 503 generates an acoustic signal transferred to a microphone 504 wherein the transferral can be described by a transfer function H(z). The signal from the microphone 504 is supplied to a beamformer unit 505 and an adaptive filter 506 for estimating the noise in the microphone signal; said microphone signal representing the total sound level in said environment, wherein the total sound level comprises both said sound output from the loudspeaker 503 and the ambient noise within the environment where the loudspeaker 503 and the microphone 504 are located (LRM system). A signal output by the beamformer unit 505 is subtracted from the signal y[k]from the DEC unit 502 in order to obtain a reference signal for the adaptive filter 506. The signal output by the adaptive filter 506 is transferred via a psychoacoustic filter 507 (e. g., an A-weighting filter) to an LPC analysis unit 508 for calculating the coefficients of the DEC unit 502.

In this combination of a beamformer (beamformer unit 505) and a unit for estimating the noise signal (adaptive filter 506) which can be designed in the manner shown in FIG. 1, for example, the voice signal - which at least in automotive applications comes from known directions - is isolated from the remaining signal components using the space/time filter effects of a beamformer, and is therefore available to the subsequent adaptive filter in addition to the actual signal required (music) as a further reference signal. The adaptive filter then excludes the signal components from the microphone signal that are correlated with the resulting reference signal. The disadvantage of this method is that the beamformer does not process the voice signal alone, but also includes noisy components that principally come from the direction in which the beamformer was directed. These noisy components adopted by the beamformer then likewise become part of the reference signal and are therefore computationally removed from the microphone signal, so that the estimated noise signal is corrupted. Normally, the degree of corruption increases with falling frequency since beamformers generally have great difficulty in obtaining adequate detector efficiency at low frequencies.

FIG. 6 illustrates a combination of a negative beamformer and an adaptive filter. In the DEC system of FIG. 6, a music signal x[k] from a music signal source 601 is fed into dynamic equalization control (DEC) unit 602 providing an equalized signal y[k] to a loudspeaker 603. The loudspeaker 603 generates an acoustic signal from the signal y[k]; said acoustic signal is transferred to a microphone 604 wherein the transferral can be described by a transfer function H(z). The signal from the microphone 604 is supplied to a negative beamformer unit 605 which is connected upstream of an adaptive filter 606 for estimating the noise in the microphone signal; said microphone signal representing the total sound level in said environment, wherein the total sound level comprises both said sound output from the loudspeaker 603 and the ambient noise within the environment (LRM system) except for voice signals. The signal y[k] from the DEC unit 602 is used as a reference signal for the adaptive filter 506. The signal output by the adaptive filter 606 is transferred via a psychoacoustic filter 607 (e. g., an A-weighting filter) to an LPC analysis unit 608 for calculating the coefficients of the DEC unit 602.

In the system of FIG. 6 the voice signal component in the microphone signal is suppressed using a so-called negative beamformer. Here the expression "negative" means that the beamformer pointing in the direction of the speaker has a spatial zero position and therefore does not detect signals from this direction that are dominated by the voice signal. The beamformer can be implemented either in software (beamforming filter) or hardware (for example, pressure gradient microphones, pipe slit microphones, etc.). The problem with this system is that what is actually of interest is the noise situation precisely where the driver, for example, is sitting or in the vicinity of the driver's head, and that signals from exactly this direction are not permitted to be excluded for this reason. However, this is what happens when using a beamformer since it does not isolate the signals in any way, but rather suppresses all signals coming from the direction in which the spatial zero point was positioned.

If the microphone signal is converted to the spectral range, an individual voice activity detector (VAD) can be applied for each spectral line, similar to individual DVC systems. Using the VAD, the voice signal component can be simply computationally removed from the microphone signal. If the VAD output signal (which corresponds to the estimated noise signal) is then again reconverted to the time domain, we know that it can be reused for the LPC analysis. Such combination between time and spectral processing are shown in FIG. 7.

In the system of FIG. 7, the voice signal component in the microphone signal is suppressed in the frequency domain. A music signal x[k] from a music signal source 701 is fed into dynamic equalization control (DEC) unit 702 providing an equalized signal y[k] to a loudspeaker 703. The loudspeaker 703 generates an acoustic signal from the signal y[k]; said acoustic signal is transferred to a microphone 704 wherein the transferral can be described by a transfer function H(z). The signal from the microphone 704 is supplied to an adaptive filter 706 for estimating the noise in the microphone signal; said microphone signal representing the total sound level in said environment, wherein the total sound level comprises both said sound output from the loudspeaker 703 and the ambient noise within the environment (LRM system). The signal y[k] from the DEC unit 702 is used as a reference signal for the adaptive filter 706. The signal output by the adaptive filter 706 is transferred via a spectral voice suppression unit 709 and a psychoacoustic filter 707 (e. g., an A-weighting filter) to an LPC analysis unit 708 for calculating the coefficients of the DEC unit 702. The voice suppression unit 709 comprises a Fast-Fourier transform (FFT) unit 710 for transforming from the time domain into the frequency domain. In a mean calculation unit 711 the signals in the frequency domain are averaged and supplied to an voice activity decoder (VAD) 712 for detecting spectral voice signal components in the signals in the frequency domain from the FFT unit 710. The signals from the VAD 712 are subsequently transformed back from the frequency domain into the time domain for being supplied to the psychoacoustic filter 707.

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.