EP1261961B1

EP1261961B1 - Methods and systems for noise reduction for spatially displaced signal sources

Info

Publication number: EP1261961B1
Application number: EP01904915A
Authority: EP
Inventors: Leonid Krasny; Ali S. Khayrallah
Original assignee: Ericsson Inc
Current assignee: Ericsson Inc
Priority date: 2000-02-29
Filing date: 2001-01-18
Publication date: 2004-03-31
Anticipated expiration: 2021-01-18
Also published as: AU2001232851A1; EP1261961A1; MY134004A; DE60102571T2; CN1406372A; ATE263410T1; WO2001065540A1; DE60102571D1

Abstract

The spatial characteristics associated with the respective locations of various signal sources within the receiver environment, such as a vehicle, are used in processing received signals from the signal sources. The use of the spatial characteristics may provide improved noise reduction and further, in various embodiments of the present invention, may be used to apply a selected suppression level to one or more of the signal sources. For example, in a vehicle hands-free speech reception system, far-end feedback from a speaker in the vehicle may be substantially suppressed while speech from a driver and a passenger in the vehicle may be processed with substantially zero decibels of suppression. The suppression levels may be user selectable.

Description

BACKGROUND OF THE INVENTION

The present invention relates to noise reduction and, more particularly, to noise reduction for antenna arrays.
As mobile terminals become more commonly used for providing communications to traveling users, it is desirable to provide hand-free use of mobile terminals by drivers of vehicles, such as automobiles. While hands-free systems have been introduced to assist a vehicle driver, noise problems may be introduced by such systems. Two approaches have previously been proposed to address this problem. One is single-microphone noise reduction techniques which may utilize differences in the spectral characteristics of speech and noise. Such systems are described, for example, in S.F.Boll. "Suppression of acoustic noise in speech using spectral subtraction," IEEE Trans. on Acoustics, Speech, and Signal Processing. ASSP-27(2):113-120, April 1979; and, R.J. McAlulay and M.L. Malpass. "Speech enhancement using a soft-decision noise-suppression filter," IEEE Trans. on Acoustics, Speech, and Signal Processing, ASSP-28:137-145, 1980. However, in many situations, the speech and the noise tend to have similar spectral distributions. Under these conditions, the single-microphone echo suppression technique may not yield substantial improvement in speech intelligibility. On the other hand, the signal and the echo in a car environment are acoustical fields which typically have different spatial characteristics. The spatial separation of the speech and the echo can be exploited to reduce the noise level.
It is known that processing of such spatial signals generally requires antenna (receiver) arrays that utilize several microphones. Approaches which utilize such arrays in conjunction with signal processing have been developed and applied in other fields, such as sonar and seismic focus searching. The general technique, called "antenna array processing", may achieve effective rejection of underwater noise (ambient noise and ocean reverberation) as described, for example, in L.G. Krasny, Spatial processing of acoustic signals in a plane-parallel waveguide, Sov.Phys.Accoust., 30, 4, 495-501, 1984 and A.B. Baggeroer, W.A. Kuperman and H. Shmidt, "Matched-field processing: source localization in correlated noise as an optimum parameter estimation problem," J.Acoust.Soc.Am. 83, 571-587,1988.
The typical conventional antenna array processing algorithm can be described by the following equation in the frequency domain: where U _out(w) and U(w,r _i) are respectively the Fourier transform of the antenna processor output and the field u(t,r _i) observed at the output of the i-th antenna element with the spatial coordinates r _i and H(w;r _i) is the frequency response of the filter at the ith antenna element, which satisfies the system of equations: where g _N(w; r _i,r _k) is the spatial correlation function of the background noise, R ₀ is the spatial coordinates of the talker, N is the number of antenna elements, j = √-1, and c is the speed of sound.
FIG. 1 illustrates a conventional system 100 for performing antenna array processing. The system includes N filters 110 (filters H ₁ ,..,H _N ) which filter N signals received from microphones 120, where N = 1,2,3,.... Preferably, N spatially displaced microphones are used and the signals from each are sampled by analog to digital (A/D) converters and the N filters 110 are implemented as digital filters. The filtered results are summed in a summer 130 and the resulting sum U _out is a signal in which the background noise is generally suppressed. The circuit 140 labeled "Est.CF" estimates the noise correlation matrix and calculates the frequency responses of the filters H ₁ ,..,H _N 110 according to equation (2) above.
However, some difficulties may become apparent when this technique is applied to the noise reduction problem in a car environment. First, while the conventional antenna array processing equation (1) generally works properly in the presence of a single sound source, its efficiency typically suffers considerably in the case of multiple sound sources. For example, if the driver and passengers talk simultaneously, equation (1) generally cannot separate these sources which may lead to a significant signal distortion after array processing. Secondly, equation (1) is based on an assumption that the antenna array is located in a free-field propagation channel. However, the free-field propagation model does not take into account affects, such as waveguide sound propagation, typically found in a car cabin environment. Accordingly, there is a need for a system to reduce the noise associated with spatially displaced signal sources.
European Patent Application Publication No. EP 0411801 describes a communication system with active noise cancellation. An acoustic attenuation system is provided with various adaptive filter models enabling communication between persons in spaced zones by selectively canceling undesired noise and undesired speech on an on-line basis.

SUMMARY OF THE INVENTION

The present invention may meet this need by providing methods, systems and mobile terminals which use the spatial characteristics associated with the respective locations of various signal sources within the receiver environment, such as a vehicle, in processing received signals from the signal sources. The use of the spatial characteristics may provide improved noise reduction and further, in various embodiments of the present invention, may be used to apply a selected suppression level to one or more of the signal sources. For example, in a vehicle hands-free speech reception system, far-end feedback from a speaker in the vehicle may be substantially suppressed while speech from a driver and a passenger in the vehicle may be processed with substantially zero decibels of suppression. The suppression levels may be user selectable. These objects are achieved by the system according to claim 1 and the method according to Claim 17.
In one embodiment of the present invention, a noise reduction system is provided including a plurality off receive channels coupled to a plurality of signal inputs. Each of the receive channels including a plurality of filters that output channel component filtered signals. Each of the filters may be responsive to at least one of the plurality of signal inputs and have coefficients based on a source spatial characteristic associated with the respective receive channel. The receive channels further include a channel combiner circuit that combines the channel component filtered signals to provide a channel filtered signal. The noise suppression system further includes a constraint circuit that outputs a suppression for at least one of the channel filtered signals and an output combiner circuit responsive to the plurality of receive channels that combines the channel filtered signals. The channel combiner circuit may include a summer that receives the channel component filtered signals and a weighting filter that filters an output of the summer to provide the channel filtered signal. The weighting filter may be responsive to the constraint circuit. The constraint circuit may be responsive to a user input designating a desired suppression for at least one of the plurality of receive channels. In one embodiment, the constraint circuit outputs coefficients of the weighting filter associated with the at least one of the plurality of receive channels as the suppression for the at least one of the channel filtered signals which is associated with the at least one of the plurality of channels based on the designated desired suppression and the source spatial characteristic associated with the at least one of the plurality of channels.
In another embodiment of the present invention, the noise suppression system includes a coefficient estimation circuit that adjusts the coefficients of the plurality of filters of each of the plurality of receive channels responsive to the audio inputs. The coefficient estimation circuit may be configured to adjust the coefficients of the plurality of filters of each of the plurality of receive channels based on the source spatial characteristic associated with the respective receive channels. In one embodiment, a plurality of receivers generate the plurality of signal inputs. The coefficient estimation circuit adjusts the coefficients of the plurality of filters of each of the plurality of receive channels based on the source spatial characteristic associated with the respective receive channels using a Green function for each of the plurality of receive channels that corresponds to a propagation channel from a signal source associated with the respective channel to the plurality of receivers. The receivers may be spatially displaced microphones and the signal sources associated with the respective channels may be sound sources in a vehicle.
In a further embodiment of the present invention, a noise reduction system for a multi-source environment is provided including a plurality of filters coupled to a plurality of signal inputs. The plurality of filters output component filtered signals. Each of the filters may be responsive to at least one of the plurality of signal inputs and have coefficients based on source spatial characteristics associated with sources in the multi-source environment. The system further includes an output combiner circuit responsive to the plurality of filters that combines the component filtered signals and a coefficient estimation circuit that adjusts the coefficients of the plurality of filters based on the source spatial characteristics. The coefficient estimation circuit may include a constraint circuit that outputs a suppression associated with at least one of the sources in the multi-source environment.
In a further embodiment of the present invention, a vehicle hands-free speech reception system is provided including a plurality of spatially displaced microphones in the vehicle and a plurality of receive channels coupled to the microphones. Each of the receive channels is associated with a respective one of a plurality of spatial positions in the vehicle. Each of the receive channels includes a plurality of filters that output channel component filtered signals, each of the filters responsive to at least one of the microphones and having coefficients based on a source spatial characteristic associated with the spatial position in the vehicle associated with the respective receive channel, and a channel combiner circuit that combines the channel component filtered signals to provide a channel filtered signal. A constraint circuit outputs a suppression for at least one of the channel filtered signals and an output combiner circuit responsive to the plurality of receive channels combines the channel filtered signals. One of the plurality of spatial positions may be a passenger position and at least one of the plurality positions may be a noise source. The noise source may be a speaker and the speaker and the vehicle hands-free reception system may be a telephone system.
In a method aspect of the present invention, a method for noise reduction is provided including receiving sound signals from a plurality of displaced spatial positions at a receiver and processing the received signals through a plurality of receive channels to provide a plurality of processed signals, each of the processed signals being associated with one of the displaced spatial positions. A selected suppression is applied to at least one of the processed received signals and the suppressed at least one of the processed received signals and the other processed signals are combined.
In a further embodiment of the present a method for noise reduction is provided including receiving signals and processing the plurality of received signals through a first filter having coefficients associated with a first source spatial characteristic to provide a first filtered signal and through a second filter having coefficients associated with a second source spatial characteristic to provide a second filtered signal. The first filtered signal is processed through a third filter having coefficients associated with a selected suppression for the first source to provide a first suppressed signal and the second filtered signal is processed through a fourth filter having coefficients associated with a selected suppression for the second source to provide a second suppressed signal. The first and second suppressed signals are combined. The signals are preferably received from N spatially displaced microphones and processed through an associated one of N filters comprising the first filter and outputs of the N filters comprising the first filter are combined to provide the first filtered signal. The received signals from each of the N microphones are also processed through an associated one of N filters comprising the second filter and outputs of the N filters comprising the second filter are combined to provide the second filtered signal.
In a further embodiment of the present invention, the coefficients of the N filters comprising the first filter and the N filters comprising the second filter are estimated. The coefficients of the third filter may be estimated responsive to a constraint value associated with the first source and the coefficients of the fourth filter may be estimated responsive to a constraint value associated with the second source. The first source may be a wanted source and the step of processing the first filtered signal through a third filter may include processing the first filtered signal through the third filter wherein the third filter provides a selected suppression of about 0 decibels. The second source may be an unwanted source and the step of processing the second filtered signal through a fourth filter may include processing the second filtered signal through the fourth filter wherein the fourth filter provides a selected suppression of at least about -3 decibels.
In another embodiment of the present invention, a method for noise reduction is provided including receiving signals from N spatially displaced microphones and processing the received signals from each of the N microphones through an associated one of N filters, each of the N filters having coefficients associated with a plurality of source spatial characteristics and a selected suppression for a source associated with each of the plurality of source spatial characteristics. The processed received signals are combined. Preferably, the coefficients of the N filters are estimated responsive to constraint values associated with the sources associated with each of the plurality of source spatial characteristics.
As will further be appreciated by those of skill in the art, while described above primarily with reference to method aspects, the present invention may also be embodied as systems.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic block diagram illustrating a conventional array processing noise reduction system;
FIG. 2 is a schematic block diagram illustrating a noise reduction system according to a first embodiment of the present invention in a hands free vehicle speech reception system;
FIG. 3 is a schematic block diagram illustrating a noise reduction system according to a second embodiment of the present invention;
FIG. 4A - 4C are graphical illustrations of exemplary source signal spectra;
FIG. 5A - 5B are graphical illustrations of exemplary constraint functions;
FIG. 6 is a graphical illustration of performance for an embodiment of the present invention; and
FIG. 7 is a flowchart illustrating operations for an embodiment of the present invention.

DETAILED DESCRIPTION OF THE INVENTION

The present invention now will be described more fully hereinafter with reference to the accompanying drawings, in which preferred embodiments of the invention are shown. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. As will be appreciated by those of skill in the art, the present invention may be embodied as methods or devices. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment or an embodiment combining software and hardware aspects.
Operations of the present invention may be described by modeling the signal field in a car as a superposition of the signals from M users and background noise (road noise, wind noise, engine noise). In addition, assume that the microphones in the receiving antenna array are placed into the car with an arbitrary geometry defining the positions of the microphones. When a mixture of the signals and background noise are incident on the array, the Fourier transform U(w,r _i) of the field u(t,r _i) received by the i-th array element has the form: where S _m(w) is the spectrum of the signal from the m-th user, G(w,r _i,R _m) is the Green function which describes a propagation channel from the m-th user (or signal (sound) source) with the spatial coordinates R _m to the antenna array, and N(w,r _i) is the Fourier transform of the noise field.
A preferred embodiment of the present invention may be described as follows: where U _out (w) is the Fourier transform of the antenna processor output, and
Equation (4) describes a multichannel system which includes M spatial channels {U ₁(w),..,U _M(w)}. The frequency responses of the filters H(w;r _i,R _m) for each of these channels are matched with the spatial characteristics of the signal from the m-th user and the background noise and satisfy the following equation: where g $\binom{-1}{N}$ (w;r _i,r _k) denotes elements of the matrix g $\binom{-1}{N}$ which is the inverse of the noise spatial correlation function g _N(w;r _i,r _k).
Therefore, in contrast to the conventional approach described in equation (1), the noise reduction system described by equation (4) utilizes determinate information about the spatial characteristic of the source signal field in a vehicle. The array processing in the m-th spatial channel may be optimized to detect a signal from the m-th source against the background noise.
If the noise spatial correlation function g _N(w;r _i,r _k) is a priori unknown, it may still be determinate as the inverse correlation matrix g $\binom{-1}{N}$ (w;r _i,r _k) can be estimated by using the adaptive algorithm, for instance: where g $\binom{-1}{N (n)}$ (w;r _i,r _k) is an estimate of the inverse noise correlation matrix g $\binom{-1}{N}$ (w;r _i,r _k) at the n-th iteration, m _g is a convergence factor, and the functions D ⁽ ⁿ ⁾(w,r _i) satisfy the following equation: As seen in the equations, the functions D ⁽ ⁿ ⁾(w,r _i) at the n-th frame are calculated using the inverse noise correlation matrix at the previous (n-1)-th frame where a frame may be selected as a determined time interval and/or number of samples of the received signal from the microphones by associated analog to digital (A/D) converters.
The output voltages of the M spatial channels may be accumulated with the weighting functions (filters) {W ₁(w),..,W _M(w)}, which satisfy the following equation: where Ψ $\binom{-1}{mn}$ (w) denotes elements of the matrix Ψ^-1(w) which is the inverse of the matrix Ψ(w) with elements: and {B ₁(w),..,B _M(w)} may be user selectable functions. The choice of these functions depends on the desired signal processing result. For example, if clear speech is desired from all M, sources, the functions B ₁(w),..,B _M(w) may be chosen as $B_{i} (w) ≡ 1, i ∈ [1, M]$ where the 1 values in one embodiment provide a suppression of 0 decibels (dB).
If the signal from some k-th user (source) is unwanted the functions B ₁(w),..,B _M(w) may be set as: $B_{i} (w) = 1, if i ≠ k, i ∈ [1, M], and$ $B_{i} (w) = 0, if i = k .$ B _i(w)=0 theoretically provides substantially complete suppression (cancellation). Intermediate values may be provided for other levels of suppression. In one embodiment B _i(w)=.1 corresponds to -10dB of suppression, B _i(w)=.5 corresponds to -3dB of suppression and so on.
FIG. 2 is a schematic block diagram illustrating an embodiment of the present invention. As shown in FIG.2 the noise reduction system includes M spatial channels 205a, 205b. Each channel 205a, 205b includes N filters 210 (filters Hm1,..,HmN for the m-th channel where m=1,...M) which filter N signals from microphones 220, where N = 1,2,3,... The microphones are preferably spatially displaced. The N filters 210 output channel component filtered signals. Each of the N filters 210 is responsive to at least one of the signal inputs from the microphones 220 and has coefficients based on a source spatial characteristic associated with the respective channel, for example, with a particular spatial location signal source within a vehicle, such as the driver, with which the respective channel is associated.
With reference to each of the M channels, the outputs of the filters Hm1,..,HmN at the m-th spatial channel are summed in respective channel combiner circuits, such as summers Σm 230, where m = 1,2,3..,M are the respective M channels. A summer 230 is included in each of the M channels 205a, 205b. The outputs of the summers Σ1,..,ΣM 230 are filtered through the suppression (weighting) filters W _1,.., W _M 240 which, in combination with the summers 230 provides a channel combiner circuit that outputs a suppression for the channel filtered signals from the filters 210 in the illustrated embodiment. The weighting filters 240 are responsive to a constraint circuit 270 to provide a desired suppression to each of the spatial location signal sources. The filtered results are summed in a summer (channel output combiner circuit) 250 and the resulting sum is a signal U _out in which the background noise may be suppressed and signals from unwanted sources may be canceled (substantially fully suppressed).
A coefficient estimation circuit 260 ("Est.CF") estimates the noise correlation matrix and calculates the frequency responses of the filters Hm1,..,HmN (m= 1,2,..,M) 210. In a preferred embodiment, the coefficients are calculated according to equation (6) and are updated responsive to the signal inputs from the microphones 220.
A constraint circuit 270 calculates the frequency responses of the filters W _1,.., W_M 240 (by providing the filter coefficients), preferably according to equation (8). The constraint circuit 270 may also generate outputs based on constraint functions B ₁(w),..,B _M(w). As mentioned above, choice of these functions depends on the goal of the circuit, such as, to keep clear speech from all users or to suppress signals from some unwanted signal sources. The constraint functions may be user selectable inputs designating a desired suppression for each of the receive channels (and thereby the associated spatial location signal source such as a driver, a passenger or a speaker). As shown in FIG. 2 and as can be seen from equation (8), the constraint circuit 270 receives the frequency responses of the filters 210 as inputs.
An alternative embodiment of the present invention is illustrated in the block diagram of FIG. 3. N filters 310 (filters H01,..,H0N) are provided which filter N digitally sampled signals received from N microphones 320. The filters 310 are responsive to the signals from the microphones 320 and have coefficients defining their frequency response based on source spatial characteristics associated with sources in the multi-user environment of the system as, for example, is shown in the embodiment described by equation (12) below. The component filtered signals from the filters 310 are combined in an output combiner circuit 350 (a summer in the illustrated embodiment) and the resulting sum is a signal U _out in which the background noise may be selectively suppressed.
In contrast to the filters H1,..,H _N 110 for the conventional array processing system in FIG. 1, the frequency responses of the filters H01,..,H0N are preferably described by the following equation: where the functions W _m(w) satisfy the system of equations (8) and the functions H(w;r _i,R _m) satisfy the system of equations (6). In other words, each filter H0i 310 may be viewed as M filters W _1,.., W _M with the frequency responses W _m (w) and M filters H _mi,.., H _mi with the frequency responses H(w;r _i, R _m) (m = 1,.., M). Using this model, the coefficient estimation circuit 360 in the embodiment of FIG. 3 estimates the noise correlation matrix and calculates the frequency responses of the filters Hm1,..,HmN (m = 1,..,M) according to the equation (6). The constraint circuit 370 calculates the frequency responses of the filters W ₁,.., W _M according to equation (8) using constraint functions B ₁(w),..,B _M(w) to select signal suppression levels for wanted and unwanted signal sources. The coefficient estimation circuit 360 and the constraint circuit 370 may be combined in constrained coefficient estimation circuit 375.
The present invention will now be further described by an example where there is just one source (a driver) in the vehicle and we want to keep the driver's speech clear. Let us assume also that we would like to suppress echo from a far-end speaker which is to be suppressed during hands-free communication. In this example M = 2 may be selected. Therefore, the system may include two spatial channels U ₁(w) and U ₂(w). The frequency responses of the filters H(w;r _i,R ₁) at the first channel may be matched with the spatial coordinates R ₁ of the driver, and the frequency responses of the filters H(w;r _i,R ₂) at the second channel may be matched with the spatial coordinates R ₂ of a loudspeaker which creates echo signal.
The functions B ₁(w) and B ₂(w) are chosen according to equations $B_{1} (w) = 1, B_{2} (w) = 0.$
A further example would be when there are two user sources (a driver and one passenger in a vehicle and we want to keep clear speech from both of them. Let us assume again a far-end signal echo is to be suppressed. In this case, we may choose M = 3. Therefore, the system may include three spatial channels U ₁(w), U ₂(w) and U ₃(w). The frequency responses of the filters H(w;r _i,R ₁) at the first channel may be matched with the spatial coordinates R₁ of the driver, the frequency responses of the filters H(w;r _i,R ₂) at the second channel may be matched with spatial coordinates R₂ of a loudspeaker generating the far-end signal, and the frequency responses of the filters H(w;r _i,R ₃) at the third channel may be matched with the spatial coordinates R₃ of the passenger.
The functions B ₁(w), B ₂(w) and B ₃(w) are chosen in this simple example according to equations: $B_{1} (w) = 1, B_{2} (w) = 0 and B_{3} (w) = 1.$ It is further to be understood that the functions B ₁(w) may be user selectable and, thus, changed by a driver or passenger in a vehicle.
More generally, the functions B₁(w) ... B_M(w) can be chosen to produce a desired overall effect. For example, there is a compromise that may be achieved among several goals. This may be illustrated by a further example.
Consider a case where source 1 is a desired speech signal, source 2 is a fan on the dashboard of a car that needs to be suppressed and the ambient noise is a mix of road noise and engine noise. The spectra S₁(w) and S₂(w) are illustrated in FIGs. 4A through 4C for purposes of illustrating this example as well as the spectrum of the noise S_noise(w). In choosing B₁(w), the goal may be to conserve source 1 while suppressing the noise. In choosing B₂(w) it is preferable to attenuate source 2. Illustrative resulting choices for this example are shown in FIGs. 5A and 5B. It is to be understood that the spectra and constraints are illustrated in a simplified manner for purposes of this example to illustrate the flexibility provided through choosing the constraint functions to reflect the spectra of desired and undesired signals.
A computer simulation of a system according to an embodiment of the present invention is shown in FIG. 6, which illustrates the output signal-to-noise ratio as a function of frequency. Solid lines correspond to an embodiment of the present invention described by equation (4), and dashed lines correspond a conventional system based on equation (1). The simulations are based on 4-element antenna array and two user sources.
For this simulation, the present invention allows substantial (25-30 dB) attenuation of the noise field and unwanted signal without substantial suppression and/or degradation of the target (desired) signal.
Operations of the present invention will now be described with respect to the flowchart illustration of FIG. 7. It will be understood that each block of the flowchart illustrations and the block diagram illustrations of FIGs. 2 and 3, and combinations of blocks in the flowchart illustrations and the block diagram illustrations, can be implemented by computer program instructions. These program instructions may be provided to a processor to produce a machine, such that the instructions which execute on the processor create means for implementing the functions specified in the flowchart and block diagram block or blocks. The computer program instructions may be executed by a processor to cause a series of operational steps to be performed by the processor to produce a computer implemented process such that the instructions which execute on the processor provide steps for implementing the functions specified in the flowchart and block diagram block or blocks.
Accordingly, blocks of the flowchart illustrations and the block diagrams support combinations of means for performing the specified functions, combinations of steps for performing the specified functions and program instruction means for performing the specified functions. It will also be understood that each block of the flowchart illustrations and block diagrams, and combinations of blocks in the flowchart illustrations and block diagrams, can be implemented by special purpose hardware-based systems which perform the specified functions or steps, or combinations of special purpose hardware and computer instructions. For example filters 210, 310, coefficient estimation circuits 260, 360 and constraint circuits 270, 370 may all be implemented as code executing on a processor, as custom chips or as a combination of the above.
Operation according to embodiments of the present invention will now be described with reference to the flowchart illustration of FIG. 7. Operations begin at block 500 with determination of a desired suppression for each of a plurality of spatial location signal sources. For example, for a hands-free speech reception system in a vehicle, a spatial location associated with a driver and a passenger respectively may be provided a desired suppression of substantially zero decibels. A spatial location associated with a speaker outputting the far-end speech signal into the vehicle compartment may be associated with a desired suppression of about minus three decibels or, preferably, with substantially complete suppression providing effective cancellation of far-end echo signals. For embodiments of the present invention including separate channels associated with each spatial location signal source, as illustrated, for example in FIG. 2, one of the channels may be associated with each of the spatial location signal sources (block 505).
Signals are received from the spatial location signal sources (block 510). Preferably, the signals are received at N spatially displaced microphones providing an antenna array. N filters H(w) are preferably provided in each channel with one of the filters being associated with each of the N microphone sources. Coefficients are estimated for each of the filters H(w) and for the suppression (weighting) filters W(w) associated with each channel (block 515). It is further to be understood that, for embodiments such as that illustrated in FIG. 3, the estimation of the coefficients H(w) and W(w) are combined and result in generation of the coefficients of the filters 310 as shown in FIG. 3.
The received signals are then processed through the receive channels to provide a plurality of process filtered signals, each of which is associated with one of the displaced spatial location signal sources as will now be described with reference to blocks 520 through 525. For the illustrated embodiment of FIG. 7 which may be implemented, for example, in the system illustrated in FIG. 2, the received signals are processed through associated ones of the N filters corresponding to the respective N microphones in each of the plurality of channels (block 520). The outputs of the filters within each channel are combined (block 525). For the illustrated embodiment shown in FIG. 7, a selected suppression is applied to the received signals by processing the combined outputs from the N filters H(w) of each channel through a suppression filter W(w) for each channel (block 530). The suppression filters W(w) have coefficients associated with a selected suppression as determined at block 500. The selected suppressions may be user selectable or may be otherwise set.
It is to be further understood that, in other embodiments of the present invention, the filter processing and suppression responsive to a spatial location associated with each signal source may be combined into a single composite filtering step, for example as with the illustrated embodiment of FIG. 3. Accordingly, operations at block 520 through 530 described above may be carried out by processing the signal from each of the end microphones through an associated filter having coefficients based upon signal source spatial characteristics and a desired suppression for each of the displaced spatial location signal sources. In either case, the outputs of the respective filters acting as suppression filters may then be combined to generate a signal (block 535).
In the drawings and specification, there have been disclosed typical preferred embodiments of the invention and, although specific terms are employed, they are used in a generic and descriptive sense only and not for purposes of limitation, the scope of the invention being set forth in the following claims.

Claims

A noise reduction system comprising:
a plurality of receive channels (205a, 205b ...) coupled to a plurality of signal inputs (220), each of the plurality of receive channels comprising,
a plurality of filters (210) that output channel component filtered signals, each of the plurality of filters responsive to at least one of the plurality of signal inputs and having coefficients based on a source spatial characteristic associated with the respective receive channel, and

a channel combiner circuit (230) that combines the channel component filtered signals to provide a channel filtered signal;
and

an output combiner circuit (250) responsive to the plurality of receive channels that combines the channel filtered signals.
The system according to Claim 1 wherein the system further comprises a constraint circuit that outputs a suppression for at least one of the channel filtered signals.
The system according to Claim 2 wherein the channel combiner circuit further comprises:
a summer that receives the channel component filtered signals; and

a weighting filter that filters an output of the summer to provide the channel filtered signal.
The system according to Claim 3 wherein the weighting filter is responsive to the constraint circuit.
The system according to Claim 4 wherein the constraint circuit is responsive to a user input designating a desired suppression for at least one of the plurality of receive channels.
The system according to Claim 5 wherein the constraint circuit outputs coefficients of the weighting filter associated with the at least one of the plurality of receive channels as the suppression for the at least one of the channel filtered signals which is associated with the at least one of the plurality of channels based on the designated desired suppression and the source spatial characteristic associated with the at least one of the plurality of channels.
The system according to any of claims 2-6 wherein the constraint circuit outputs a suppression for at least one of the channel filtered signals based on a frequency spectrum of the signal source associated with the receive channel providing the at least one of the channel filtered signals.
The system according to any of claims 1-7 wherein the system further comprises a coefficient estimation circuit that adjusts the coefficients of the plurality of filters of each of the plurality of receive channels responsive to the signal inputs.
The system according to Claim 8 wherein the coefficient estimation circuit is further configured to adjust the coefficients of the plurality of filters of each of the plurality of receive channels based on the source spatial characteristic associated with the respective receive channels.
The system according to Claim 9 further comprising a plurality of receivers that generate the plurality of signal inputs and wherein the coefficient estimation circuit adjusts the coefficients of the plurality of filters of each of the plurality of receive channels based on the source spatial characteristic associated with the respective receive channels using a Green function for each of the plurality of receive channels that corresponds to a propagation channel from a signal source associated with the respective channel to the plurality of receivers.
The system according to Claim 10 wherein the receivers are spatially displaced microphones and the at least one signal source associated with the respective channels is at least one sound source.
The system according to Claim 11 wherein the spatially displaced microphones and the at least one sound source are in a vehicle.
A receive channel according to any of claims 1-12.
The system of Claim 1 wherein the system is a hands-free speech reception system in a vehicle, the system further comprising:
a plurality of spatially displaced microphones in the vehicle, each of the microphones coupled to the plurality of receive channels to provide the plurality of signal inputs; and
wherein each of the receive channels is associated with a respective one of a plurality of spatial positions in the vehicle and wherein each of the source spatial characteristics associated with the respective receive channels are associated with a respective one of a plurality of spatial positions in the vehicle.
The system according to Claim 14 wherein at least one of the plurality of spatial positions is a passenger position and at least one of the plurality positions is a noise source.
The system according to Claim 15 wherein the noise source is a speaker and wherein the speaker and the hands-free reception system comprise a telephone system.
A method for noise reduction comprising:
receiving a plurality of received signals (220);

processing the plurality of received signals through a first filter (210, 230 in 205a) having coefficients associated with a first source spatial characteristic to provide a first filtered signal;

processing the plurality of received signals through a second filter (210, 230 in 205b) having coefficients associated with a second source spatial characteristic to provide a second filtered signal;

processing the first filtered signal through a third filter (240 in 205a) having coefficients associated with a selected suppression for the first source to provide a first suppressed signal;

processing the second filtered signal through a fourth filter (240 in 205b) having coefficients associated with a selected suppression for the second source to provide a second suppressed signal; and

combining (250) the first and second suppressed signals.
The method according to Claim 17 wherein receiving a plurality of received signals comprises receiving the plurality of received signals from N spatially displaced microphones and wherein processing the plurality of received signals through a first filter having coefficients associated with a first source spatial characteristic to provide a first filtered signal comprises processing the plurality of received signals from each of the N microphones through an associated one of N filters comprising the first filter and combining outputs of the N filters comprising the first filter to provide the first filtered signal and wherein processing the plurality of received signals through a second filter having coefficients associated with a second source spatial characteristic to provide a second filtered signal comprises processing the plurality of received signals from each of the N microphones through an associated one of N filters comprising the second filter and combining outputs of the N filters comprising the second filter to provide the second filtered signal.
The method according to Claim 18 further comprising estimating the coefficients of the N filters comprising the first filter and the N filters comprising the second filter.
The method according to Claim 19 further comprising estimating the coefficients of the third filter responsive to a constraint value associated with the first source and estimating the coefficients of the fourth filter responsive to a constraint value associated with the second source.
The method according to Claim 19 wherein estimating the coefficients of the N filters comprising the first filter and the N filters comprising the second filter comprises estimating the coefficients based on the equation where g $\binom{-1}{N}$ (w;r _i,r _k) the matrix g $\binom{-1}{N}$ which is an inverse of the noise spatial correlation function g _N(w;r _i,r _k) and where G(w,r _k,R _m) is a Green function which describes a propagation channel from a m-th signal source with the spatial coordinates R _m.
The method according to Claim 19 wherein the first source is a wanted source and processing the first filtered signal through a third filter having coefficients associated with a selected suppression for the first source to provide a first suppressed signal comprises processing the first filtered signal through the third filter wherein the selected suppression provided by the third filter is about 0 decibels and wherein the second source is an unwanted source and processing the second filtered signal through a fourth filter having coefficients associated with a selected suppression for the second source to provide a second suppressed signal comprises processing the second filtered signal through the fourth filter wherein the selected suppression provided by the fourth filter is at least about -3 decibels.