Audio detection device, vehicle with an audio detection device and method for controlling a hands-free device

Abstract

An audio recording device for multiple participants comprises a first and a second microphone unit (107a, b) with anisotropic directional characteristics, directed towards a first and second spatial area (106a, b), respectively. The spatial areas are distinct to predominantly capture one of two different acoustic source spectra, the sources of which are located in the first and second spatial areas. Each microphone unit (107a, b) emits an output spectrum containing components of both acoustic source spectra (S1(ω), S2(ω)). The hands-free device includes a first compensation stage 201, which receives the output spectrum of the first microphone unit (107a) as the usable spectrum and the output spectrum of the second microphone unit (107b) as the reference spectrum. The compensation stage is configured to reduce components in the usable spectrum that correspond to the reference spectrum.

Description

Area

[0001] The present invention relates to an audio detection device, a vehicle with an audio detection device and a method for controlling a hands-free device. background

[0002] In this context, the term "signal" is used for time-dependent signals which, when transformed into the frequency domain (Fourier), are called "spectrums." The two approaches are equivalent for describing the physical effects and technical consequences, although the frequency domain approach is simpler for mathematical description, as will be explained below.

[0003] Hands-free audio systems allow users to make phone calls without holding a microphone or phone, or wearing a headset. For this reason, hands-free systems are used in a variety of applications. They always serve to receive and, if necessary, transmit an audio signal from the user. This allows the user to move freely and keep their hands free for other tasks. Hands-free systems are particularly common in vehicles, allowing drivers to make phone calls while keeping their hands on the steering wheel. Furthermore, hands-free systems can also be part of a telephone or video conferencing system.

[0004] In all applications, the problem arises that an audio capture or hands-free device receives not only the audio signal of the respective user, but also ambient noise and / or audio signals from other people within range of the device. This problem is conventionally addressed by using microphones with anisotropic polar patterns, so that the microphone receives acoustic source signals from a specific area of ​​the room with higher sensitivity than acoustic signals from other directions. Therefore, microphones with a polar pattern oriented towards the driver's position, such as microphone arrays, are frequently used in motor vehicles.Microphone arrays enable spatial selectivity with respect to the driver's position and direction using a beamforming technology. This allows for attenuation of acoustic signals reaching the microphone array from other directions. In a situation where the driver is on a phone call while the passenger is speaking, the problem arises that, due to limited spatial selectivity, the microphone unit assigned to the driver cannot completely suppress the passenger's voice. Therefore, the passenger's voice is also picked up by the driver's microphone array, albeit at a lower level than the driver's. This effect, referred to as passenger crosstalk, can interfere with the other participants in the driver's phone conversation.This crosstalk is not limited to the front passenger, but can also be caused by other passengers who, for example, are having a conversation with the front passenger or another phone call from the back seat. Generally, this crosstalk is undesirable due to its disruptive nature.

[0005] Conventional approaches to solving this problem include increasing the selectivity of the driver's microphone array or optimizing the microphone array to attenuate noise around the passenger's head as much as possible. However, neither approach completely eliminates crosstalk, but at best reduces it. A disadvantage of the first approach is its high cost. The second approach worsens the noise suppression of diffuse sound, such as that caused by wind, engine, or tire noise, because the microphone is optimized to minimize passenger crosstalk, thereby simultaneously reducing the attenuation of other ambient noise. Description of the invention

[0006] It is therefore an object of the present invention to propose an audio detection device which at least partially overcomes the disadvantages described above. In particular, it is an object of the present invention to propose an audio detection device which completely or partially removes unwanted audio signals from an output signal.

[0007] This problem is solved according to a first aspect of the invention by an audio acquisition device comprising a first and a second microphone unit. The first microphone unit is configured to primarily acquire a first source spectrum from a first spatial region superimposed by a second source spectrum from a second spatial region and output it as a spectrum. The second microphone unit is configured to primarily acquire a second source spectrum from a second spatial region and output it as a spectrum. The first and second source spectrums, as well as the first and second spatial regions, differ from each other. The audio acquisition device includes a first compensation stage which receives the output spectrum of the first microphone unit as the useful spectrum and the output spectrum of the second microphone unit as a reference spectrum.The compensation stage is designed to generate a filtered signal from the reference spectrum equivalent reference signal, which replicates the reference signal in a form as it is contained in the useful spectrum equivalent signal, and to subtract the filtered signal from the useful signal.

[0008] Preferably, the compensation stage includes an adaptive filter. The adaptive filter adjusts the filtered signal to compensate for changes in how the reference signal is contained within the useful signal.

[0009] In one embodiment, the audio acquisition device is used in a hands-free system in a vehicle, where the first microphone unit is assigned to the driver and the second microphone unit to the passenger. The acoustic spectrum originating, for example, from the passenger, which is contained in the usable spectrum of the driver's microphone unit due to crosstalk, is attenuated by the directional characteristics of the microphone unit compared to the acoustic spectrum component emitted by the driver. The unwanted spectrum component is further reduced by a compensation unit. This has the advantage that no disproportionately large effort needs to be made to the directional characteristics of the microphone unit in order to significantly reduce the unwanted spectrum component relative to the desired usable spectrum component.

[0010] According to an advantageous embodiment, the hands-free device has a second compensation stage assigned to the second microphone unit, so that both microphone units are each equipped with a compensation stage. Each compensation stage is alternately fed one of the output spectra of the microphone units as the usable spectrum and the other output spectrum as the reference spectrum.

[0011] In this preferred embodiment, the arrangement is symmetrical in the sense that the desired usable spectrum can come from the driver or the passenger, with the other spectrum being recorded as a reference spectrum and its components being removed from the usable spectrum as completely as possible.

[0012] In practical embodiments of the microphone device, both acoustic source signals are included in filtered form in the output signal of each microphone unit, with one of the two source spectra being reduced in magnitude relative to the other source spectrum in the output spectrum of the microphone units by a damping factor. The damping factor is usually less than 0.9 and preferably lies in the range of 0.05 to 0.5.

[0013] Damping factors of this magnitude can be achieved without particular difficulty using directional microphones or microphone arrays.

[0014] In a particularly preferred embodiment, each compensation stage is configured to reduce one of the two source spectra in the output spectrum of the compensation stage by the square of the attenuation factor relative to the other source spectrum.

[0015] The quadratic dependence on the damping factor results in very effective attenuation of unwanted spectrum components in the usable spectrum of the hands-free device, without having to exert a high degree of effort with regard to the directional characteristics of the microphone units.

[0016] According to a second aspect, the invention proposes a vehicle with an interior and an audio detection device according to the first aspect of the invention. The first microphone unit preferably receives acoustic source spectra from a spatial area of ​​the driver's seat, and the second microphone unit preferably receives acoustic source spectra from a spatial area of ​​the passenger's seat.

[0017] With a vehicle equipped with an audio detection device according to the first aspect of the invention, all the advantages mentioned in connection with the audio detection device are realized.

[0018] Finally, according to a third aspect of the invention, a method for acquiring an acoustic source spectrum with an audio acquisition device according to the first aspect of the invention is proposed. The method comprises - Capturing acoustic source spectra with the first microphone unit, which are then delivered as a usable spectrum to the compensation stage, - Capturing acoustic source spectra with the second microphone unit, which are then passed on to the compensation stage as a reference spectrum, - Filtering the reference signal equivalent to the reference spectrum with a filter, and - Subtracting a filtered version of the reference signal from the useful signal of the first microphone unit to generate an output signal for the compensation stage.

[0019] Especially in a vehicle where the driver and passengers assume a relatively fixed position while speaking, it is possible to determine a suitable filter function through measurements. According to a first embodiment, the filter function remains unchanged during the acquisition of the acoustic source spectra.

[0020] In a continuing education course, the procedure includes the optional step - Performing adaptive filtering to generate the output signal of the compensation stage.

[0021] In a suitable embodiment, the method comprises the step - Subtracting filtered versions of several reference signals from the useful signal of the first microphone unit.

[0022] The inventive method and its optional further developments have the advantage that unwanted spectrum components in a useful spectrum are not only reduced by means of the directional characteristic of the microphone unit involved, but are also further attenuated by means of the compensation stage. Brief description of the characters

[0023] The invention is explained in more detail below using one embodiment as an example, with reference to the accompanying figures. All figures are purely schematic and not to scale. They show: Fig. 1 a schematic representation of signal transmission in a hands-free device in a vehicle, Fig. 2 a schematic block diagram of an embodiment of an audio recording device according to the invention, Fig. 3 an embodiment of an audio recording device according to the invention with two compensation stages, Fig. 4 an embodiment of an audio detection device according to the invention with a multi-channel compensation stage, and Fig. 5 a schematic flowchart of a method for capturing an acoustic source spectrum with an audio capture device according to the invention.

[0024] Identical or similar elements in the figures are marked with the same or similar reference symbols. Description of exemplary implementations

[0025] Fig. Figure 1 schematically shows a first embodiment of a hands-free device 100 in a vehicle, the interior of which is indicated by a frame 101. A first and a second microphone array 102a, b are arranged in the interior 101, each comprising several individual microphones 103a, b. In the following, a distinction is made in some places between the driver's microphone array 102a and the passenger's microphone array 102b. The microphone arrays 102a, b are each connected to a beamformer 104a, b, which processes the signals from the individual microphones 103a, b such that the sensitivity of the microphone arrays 102a, b is spatially selective. The areas of greatest sensitivity of the microphone arrays 102a, b are shown in Fig. Figure 1 is symbolically represented as trapezoidal areas 106a, b. The arrangement consisting of a microphone array 102a (102b) and a beamformer 103a (103b) is hereinafter also referred to as a microphone unit 107a (107b). Where expedient, a distinction is also made between a driver microphone unit 107a and a passenger microphone unit 107b. In another embodiment, at least one microphone unit 107a, b is designed as a single microphone with a directional characteristic, wherein the greatest sensitivity of the single microphone(s) is directed towards the driver or passenger.

[0026] Specifically, the driver microphone array 102a preferably receives an acoustic source signal s1(t) from a driver 108a and outputs it as a signal x1(t) by the microphone unit 107a. The output signal x1(t) can, for example, be output as a speech signal in a telephone system. The same applies to an acoustic source signal s2(t) from a passenger 108b, which is preferably received by the passenger microphone unit 107b and output as a signal x2(t). In this example, the driver and passenger are the sources of the acoustic source signals s1(t) and s2(t), respectively. The acoustic signal s1(t) is preferably received by the driver microphone unit 107a, but also, in an attenuated form, by the passenger microphone unit 107b. The same applies to the acoustic source signal s2(t), with the driver and passenger microphone units 107a and 107b switching roles.

[0027] In the following, instead of the time-dependent signals, their corresponding Fourier-transformed frequency-dependent spectra are considered, as this simplifies the mathematical representation. The time-dependent signals are denoted by lowercase letters, while the frequency-dependent spectra are denoted by uppercase letters. In particular, the angular frequency ω is omitted as an argument for the spectra for the sake of readability wherever it is clear from the context that spectra are meant and where the relationship shown holds true for all angular frequencies.

[0028] The conversion of the acoustic source spectra S1(ω) and S2(ω) into the output spectra X1(ω) and X2(ω) can be represented using a 2 × 2 transfer matrix Ĥ(ω): (X1X2)=(H11H12H21H22)︸H⏜(S1S2) Multiplying out component by component yields X1=H11⋅S1︸X11+H12⋅S2︸X12 X2=H21⋅S1︸X21+H22⋅S2︸X22

[0029] The transfer matrix Ĥ(ω) is in Fig. 1. This is also graphically represented with arrows illustrating how and to what extent the source spectra S1(ω) and S2(ω) contribute to both output spectra X1(ω) and X2(ω). The arrow thickness symbolizes the respective extent to which the acoustic source spectra S1(ω) and S2(ω) contribute to the output spectra X1(ω) and X2(ω). In other words: The components H 11 and H 22 The components in the main diagonal of the transfer matrix Ĥ are larger in magnitude than the components H. 12 and H 21 in the off-diagonal of the transfer matrix Ĥ. The relative size of the off-diagonal compared to the main diagonal is expressed as a damping factor α, which is explained further below (see equations (10) and (11)).

[0030] From equations (2a) and (2b) and the graphical representation in Fig. 1. It follows that if the components H 12(ω) and H 21 If (ω) were equal to zero, then there would be no crosstalk from the passenger's acoustic source signal s2(t) into the driver's output signal x1(t), and correspondingly no crosstalk from the driver's acoustic source signal s1(t) into the passenger's output signal x2(t). In practice, the components H 12 (ω) and H 21 (ω) however, is greater than zero in magnitude, causing the described crosstalk. The aim of the invention is to reduce the crosstalk further than can be achieved by means of the spatial selectivity of the microphone arrays, without incurring the disadvantages associated with conventional techniques.

[0031] Fig. Figure 2 shows a schematic block diagram of an embodiment of the hands-free device 100 with microphone units 107a,b, whose output spectra X1(ω) and X2(ω) are fed to a compensation stage 201. The output signal x1(t) is fed as the useful signal to an adder 202, while the output signal x2(t) is fed as a reference signal to an adaptive filter 203. An output signal D(ω) of the adaptive filter 203 forms a second input signal for the adder 202 with the opposite sign. The output spectrum of the adder Y(ω) is fed to the adaptive filter 203 as an error spectrum E(w), whereby the adaptive filter 203 strives to minimize the magnitude of the error spectrum E(ω) by adjusting the filter coefficients. In this way, the compensation stage 201 reduces signal components in the useful signal completely or at least partially that are correlated with the reference signal.

[0032] In an alternative embodiment, the output spectra X1(ω) and X2(ω) exchange roles, so that the output spectrum X2(ω) is fed to the adder 202 as the useful spectrum and the output spectrum X1(ω) is fed to the adaptive filter 203 as the reference spectrum. The two variants are symmetrical with respect to the driver and the passenger, so that, without limiting the generality of the invention, the following description refers only to the variant in which the output spectrum X1(ω) is the useful spectrum and the output spectrum X2(ω) is the reference spectrum. In practical applications, both variants are implemented simultaneously, as shown in Fig. Figure 3 is shown schematically. The output spectrum Y1(ω) of compensation stage 201a carries the output spectrum X1 as its useful spectrum, while the output spectrum Y2(ω) of compensation stage 201b carries the output spectrum X2(ω) as its useful spectrum. For the sake of simplicity, however, the following will again refer generally to the output spectrum Y(ω).

[0033] Again with reference to Fig. 2. G(ω) is the transfer function of the adaptive filter 203, which is continuously adjusted. Initially, the adaptive filter uses an estimated transfer function.

[0034] With G(ω) as the transfer function of the adaptive filter 203, the following applies to the output spectrum of the compensation stage 201. Y(ω)=X1(ω)−G(ω)⋅X2(ω).

[0035] Substituting equations (2a, (2b) into equation (3) and rearranging, it follows that Y(ω)=(H11(ω)−G(ω)⋅H21(ω))⋅S1(ω)+(H12(ω)−G(ω)⋅H22(ω))⋅S2(ω).

[0036] In the optimal case, the influence of the source spectrum S2(ω) on the output spectrum Y(ω) of the compensation stage 201 disappears. For this to happen, the factor in front of the source spectrum S2(ω) must become zero, i.e. 0=(H12(ω)−Gopt(ω)⋅H22(ω)).

[0037] The optimal transfer function G opt (ω) of the adaptive filter is thus by Gopt(ω)=H12(ω)H22(ω) Given, in order to remove the portion of the acoustic source spectrum S2(ω) of passenger 108b from the output spectrum Y(ω) of compensation stage 201. The (optimal) output spectrum of compensation stage 201 thus takes the following form. Yopt(ω)=(H11(ω)−Gopt(ω)⋅H21(ω))⋅S1(ω)=(6)(H11(ω)−H12(ω)H22(ω)⋅H21(ω))⋅S1(ω)

[0038] Equation (7) can be transformed into Yopt(ω)=X11(ω)+ΔX1(ω) where ΔX1(ω)=−(H12(ω)H22(ω)⋅H21(ω))⋅S1(ω) This error term reduces the usable spectrum compared to a situation without crosstalk. This error term arises because, according to equation (2b), the output spectrum X2(ω) of the passenger microphone unit 107b contains a portion of the source spectrum S1(ω) of the driver signal 108a, namely H. 21 ω)·S1(ω). The spectral component H 21 ω).S1(ω) feeds into the reference spectrum for the adaptive filter 203. The adaptive filter 203 attempts to eliminate this signal component from the output signal of the compensation stage 201, which is reflected in the error term ΔX1(ω). In practice, the error term ΔX1(ω) in the output spectrum of the compensation stage 201 will manifest itself as a kind of signal artifact, which, however, is significantly smaller than the spectrum component H to be removed from the output spectrum of the driver microphone unit 107a. 12 ω)·S2(ω), which originates from the crosstalk of the source spectrum S2(ω) of the passenger 108b.

[0039] The directivity of the microphone units 107a, b leads to the following for the components of the acoustic transfer function. |H12(ω)|<α⋅|H11(ω)| |H21(ω)|<α⋅|H22(ω)| with α << 1, which intuitively means that the crosstalk signal is transmitted with a smaller amplitude than the main signal picked up by the microphone unit assigned to the respective speaker. This is immediately obvious because it is the purpose of the directional function of microphone units 107a, b. In practice, the attenuation factor is frequency-dependent because the directional characteristic of the microphone units is more pronounced for high frequencies than for low frequencies. This property is simplified and accounted for in equations (10) and (11) and subsequent equations by assuming the attenuation factor α to be a constant. This constant corresponds approximately to the value of the frequency-dependent attenuation factor at a low frequency relevant for speech intelligibility, for example, 500 Hz. A reasonable value is, for example, α = 0.1.For higher frequencies, the damping factor α is smaller, which results in an attenuation of at least 20 dB of the passenger signal compared to the driver signal in the output signal of microphone unit 107a across all frequencies relevant for speech recognition. The same applies to microphone unit 107b with respect to the driver and passenger signals, respectively. This explains the less-than signs in equations (10) and (11).

[0040] The interference component of the passenger spectrum in the output spectrum X1(ω) of microphone unit 107a can be estimated via the ratio of the magnitude spectra of the source spectra of the driver (S1(ω)) and the passenger (S2(ω)). The following applies: |X12(ω)X11(ω)|=(2)|H12(ω)H11(ω)|⋅|S2(ω)S1(ω)|≤(11)α⋅|S2(ω)S1(ω)| under the highly simplified assumption that the source spectra of the driver and the passenger are approximately equal in magnitude for all angular frequencies ω, i.e. |S2(ω)S1(ω)|=1 Does this mean that for α = 0.1, the interference component of the passenger spectrum X 12 (ω) in the output spectrum X1(ω) of the microphone unit 107a only a maximum of 1 / 10 of the usable portion of the driver spectrum X 11 (ω) constitutes.

[0041] The output spectrum Y(ω) of compensation stage 201 is the sum of the maximum usable spectrum of the driver H 11 (ω) · S1(ω) is subject to the error term ΔX1(ω). Relating the error term ΔX1(ω) to the maximum usable spectrum yields... |ΔX(ω)X11(ω)|=|H12(ω)⋅H21(ω)H11(ω)⋅H22(ω)|≤|αH11(ω)⋅αH22(ω)H11(ω)⋅H22(ω)|=α2.

[0042] This means that the damping factor α enters the ratio of the error spectrum to the useful spectrum quadratically. For example, with a damping factor of α = 0.1, the error spectrum in the output spectrum Y(ω) of the compensation stage is thus attenuated by 40 dB. This means that the damping factor α of the source spectra S1 and S2 does not need to be particularly high to achieve good attenuation of the unwanted source spectrum with the adaptive filter. However, it is essential that α < 1 so that attenuation of the unwanted spectra is achieved with the device consisting of the microphone unit and the attenuation unit. Unlike in the prior art, in this case the unwanted source spectrum, for example S2, is attenuated by an interaction between the directional characteristic of the microphone unit 107a and the compensation stage 201, which includes the adaptive filter 203.

[0043] In Fig. Figure 4 shows a further embodiment of a hands-free device 400 according to the invention, comprising four microphone units 107a-d, each consisting of a microphone array 102a-d with individual microphones 103a-d and a beamformer 104a-d. The microphone units 107a-d are configured to capture the acoustic source spectra of the driver, the front passenger, and passengers in the rear seats. Assuming that the acoustic source spectrum of the driver is to be captured by microphone unit 107a, it is advantageous to attenuate the acoustic source spectra of all other passengers. For this purpose, the output spectra X2(ω), X3(ω), and X4(ω) of the microphone units 107b-d are fed as reference spectra to a compensation stage 401, which has an adder 402 and an adaptive filter 403 with multiple reference channels.The compensation stage 401 and the adaptive filter 403 are set up in such a way that the components of the three reference spectra of the microphone units 107b-d are essentially removed from the output spectrum of the compensation stage 401, as already described in connection with . Fig. 2 was described.

[0044] Fig. Figure 5 shows a schematic flowchart for acquiring an acoustic source spectrum with a hands-free device 100. First, acoustic source spectra are acquired with the first microphone unit 107a and output as the useful spectrum to the compensation stage (step ST1). In the next step ST2, acoustic source spectra are acquired with the second microphone unit 107b and output as the reference spectrum to the compensation stage. In step ST3, the filtered reference signal is subtracted from the useful signal of the first microphone unit to generate an output signal for the compensation stage. Since the acoustic paths in a vehicle are relatively deterministic because the head positions of the driver and passengers vary little, an a priori measurement of the transfer functions could also be performed to obtain the necessary filter coefficients for filtering the reference signal.

[0045] Optionally, in step ST4, adaptive filtering is performed to generate the output spectrum of the compensation stage. According to a further development of the method, in step ST5, several reference spectra can be subtracted from the useful spectrum of the first microphone unit. Since steps ST4 and ST5 are only optional, they are omitted in Fig. 5 shown with dashed lines.

[0046] In another embodiment, not shown in the figures, the audio capture device is part of a stage audio system. During a musical performance, at least one singer typically uses a first microphone. This microphone also unintentionally picks up the signal from, for example, a drum kit, other musical instruments, or other singers on stage. However, in the following, only the example of the drum kit will be considered as representative of all other sound sources. The source signal of the drum kit is captured by a separate second microphone. For a sound engineer creating an output mix of all source signals, it is desirable to be able to adjust, for example, the singer's volume independently of all other output signals at the mixing console.In conventional audio systems, however, crosstalk from other sound sources causes the drums to become louder when the sound engineer increases the level of the first microphones that capture the singer's output spectrum. With the audio capture device according to the invention, it is possible to remove the output signal of the drums (as well as other sound sources) from the singer's output signal in the same way that the output signal of a passenger is removed from the output signal of a driver in a hands-free system in a vehicle. Specifically, any unwanted audio signal in a useful signal is removed by a compensation stage, as described above with reference to a hands-free system.

[0047] Similarly, the audio recording device can also be used in a recording studio, where each singer and instrument is recorded with its own microphone setup. Using a suitably configured audio recording device and a corresponding compensation stage, it is possible to completely or at least partially remove all other audio signals from each desired signal.

[0048] Another application for the audio acquisition device according to the invention is a teleconferencing system where two participants are in the same room and each participate in the same teleconference using their own access devices. In this case, the effect occurs that when one of the two participants speaks into their microphone, their speech is also unintentionally picked up by the microphone of the other participant and fed into the teleconference. As a result, one participant hears their own speech with a slight delay on their headset. This is generally perceived as very annoying. With the audio acquisition device according to the invention, this effect can be eliminated by capturing the output spectrum of one participant as the usable spectrum in a compensation stage, while the output spectrum of the other participant is captured as the reference spectrum.The reference signal, equivalent to the reference spectrum, is filtered to recreate the reference signal as it is contained within the desired signal. This filtered reference signal is then subtracted from the desired signal, ensuring that each participant in the teleconference does not hear their own output signal, but only contributions from other teleconference participants. If more than a certain number of participants are involved in the teleconference, a correspondingly larger number of source spectra than reference spectra (or reference signals) are captured and removed from the desired signal using an appropriately designed compensation stage. Reference symbol list 100 Hands-free device 101 Interior 102a, b microphone array 103a, b Single microphone 104a, b Beamformer 106a, b Sensitivity range 107a, b microphone unit 108a Driver 108b Passenger 112 Sensitivity range 201 Compensation level 202 Adders 203 Adaptive Filter 400 Hands-free device 401 Compensation level 402 Adders 403 Adaptive Filter

Claims

[1] Audio recording device with a first and a second microphone unit (107a, b), wherein the first microphone unit (107a) is configured to primarily capture a first source spectrum (S1(ω)) from a first spatial area (106a) and superimposed by a second source spectrum (S2(ω)) from a second spatial area (106b) and output it as an output spectrum (X1(ω)), wherein the second microphone unit (107b) is configured to primarily capture a second source spectrum (S2(ω)) from a second spatial area (106b) and output it as an output spectrum (X2(ω)), where the first and second source spectrums as well as the first and second spatial regions differ from each other, wherein the audio acquisition device has a first compensation stage (201,401) which receives the output spectrum (X1(ω)) of the first microphone unit (107a) as the useful spectrum and the output spectrum (X2(ω)) of the second microphone unit (107b) as the reference spectrum, wherein the compensation stage (201,401) is set up to generate a temporary intermediate signal (equivalent to D(ω)) from the reference signal equivalent to the reference spectrum by linear filtering, which replicates the reference signal in a form as it is contained in the useful signal equivalent to the useful spectrum, and to subtract the temporary intermediate signal (equivalent to D(ω)) from the useful signal. [2] Audio recording device according to claim 1, wherein the compensation stage (201,401) comprises an adaptive filter (203,403). [3] Audio acquisition device according to claim 1 or 2, comprising a second compensation stage (201) assigned to the second microphone unit (107b), such that both microphone units are each assigned to a compensation stage and each compensation stage is alternately supplied with one of the output spectra of the microphone units (107a, b) as the useful spectrum and the other output spectrum as the reference spectrum. [4] Audio recording device according to one of the preceding claims, wherein both acoustic source spectra (S1(ω), S2(ω)) are included in the output spectrum of each microphone unit (107a,b), wherein one of the two source spectra is reduced in magnitude compared to the other source spectrum in the output spectrum of the microphone units by a damping factor (α), and wherein the damping factor is less than 0.9 and preferably in the range below 0.

5. [5] Audio recording device according to claim 4, wherein each compensation stage (201) is configured to reduce one of the two source spectra (S2 (ω)) relative to the other source spectrum (S1(ω)) in the output spectrum (X1(ω)) of the compensation stage (201) by the attenuation factor squared (α). 2 to reduce. [6] Vehicle with an interior (101) and an audio detection device according to one of the preceding claims, wherein the first microphone unit (107a) receives acoustic source spectra preferably from a spatial area (106a) of a driver's seat and the second microphone unit (107b) receives acoustic source spectra preferably from a spatial area (106b) of a passenger's seat. [7] Method for capturing an acoustic source spectrum with an audio capture device according to any one of claims 1 to 5, comprising the steps: - Acquisition (ST1) of acoustic source spectra (S1(ω), S2(ω)) with the first microphone unit (107a), which are delivered as useful spectrum (X1(ω)) to the compensation stage (201), - Acquisition (ST2) of acoustic source spectra (S1(ω), S2(ω)) with the second microphone unit (107b), which are delivered as a reference spectrum to the compensation stage, - Filtering (ST3) the reference signal equivalent to the reference spectrum with a filter (203), and - subtract (ST4) a filtered version of the reference signal (x2(t)) from the useful signal (x1(t)) of the first microphone unit to generate an output signal (y(t)) of the compensation stage (201,401). [8] Method according to claim 7, comprising the steps - Performing (ST3) adaptive filtering to generate the output signal (y(t)) of the compensation stage. [9] Method according to claim 7 or 8, comprising the steps - subtract (ST5) filtered versions of several reference signals from the useful signal of the first microphone unit.

Images