JP4514153B2 - Sound equipment - Google Patents

Description

  The present invention relates to an acoustic device that estimates an acoustic transfer function of a vehicle interior space or the like.

2. Description of the Related Art Conventionally, there has been known a voice input device that simulates an audio sound that is output from a speaker and reaches a microphone by estimating an acoustic transfer function in a vehicle interior space (see, for example, Patent Document 1). The audio input device includes an audio sound removing unit, and the audio sound removing unit estimates the acoustic transfer function of the vehicle interior space by updating the coefficient of the adaptive filter by the LMS algorithm.
Japanese Patent Laid-Open No. 2001-236090 (page 3-6, FIG. 1-7)

  By the way, in the conventional method disclosed in Patent Document 1, since the acoustic transfer function of the vehicle interior space from the speaker to the microphone is estimated, it is necessary to estimate the acoustic transfer function by the number of speakers when there are a plurality of speakers. In addition, there is a problem that the amount of processing (the number of product-sum operations) for estimating the acoustic transfer function increases, and the hardware scale increases and the cost increases. Since the audio device actually mounted on the vehicle includes at least two left and right speakers, at least twice the configuration necessary for estimating the acoustic transfer function corresponding to one speaker is required. .

  The present invention was created in view of the above points, and an object of the present invention is to provide an audio device that can reduce the processing amount, reduce the hardware scale, and realize cost reduction. There is to do.

  In order to solve the above-described problems, the acoustic device of the present invention includes a first speaker and a second speaker installed at symmetrical positions in an acoustic space, and a microphone disposed at an equal distance from the first and second speakers. And estimating the first acoustic transfer function from the first speaker to the microphone and the second acoustic transfer function from the second speaker to the microphone, whereby the first and second included in the output signal of the microphone Component extraction means for extracting a component corresponding to the output sound of the speaker, and the component extraction means includes common component extraction means for performing extraction corresponding to the common component of the first and second acoustic transfer functions, Difference component extraction means for performing extraction corresponding to the difference component of the second acoustic transfer function while paying attention to the high frequency component. When two speakers are installed at symmetrical positions in the acoustic space, and microphones are arranged at an equal distance from each of these speakers, the respective acoustic transfer functions from the two speakers to the microphones have low frequency components (low frequency components). The component) is considered to be substantially the same, and therefore, for this difference component, attention should be paid to the high frequency component (high frequency component). In this way, by dividing each acoustic transfer function from each of the two speakers to the microphone into a common component and a difference component, the processing corresponding to the low frequency component can be omitted for the difference component. The amount of processing can be reduced, and the cost can be reduced by reducing the hardware scale.

  In addition, the common component extraction unit includes a first adaptive filter that estimates an acoustic transfer function corresponding to the common component using a signal input to each of the first and second speakers as a reference signal, and extracts a difference component. The means includes a second adaptive filter for estimating an acoustic transfer function corresponding to a difference component using a signal input to the second speaker as a reference signal, and the number of taps of the second adaptive filter is determined by the first adaptive filter. It is desirable to set less than the number of taps. Specifically, the first adaptive filter described above includes N first multiplication means for multiplying a predetermined tap coefficient after delaying the input reference signal, and common component extraction means from the output signal of the microphone. And first coefficient update processing means for updating the value of the tap coefficient so that the power of the error signal obtained by subtracting the outputs of the difference component extraction means is minimized, and the second adaptive filter receives the input reference Less than N second multiplication means for multiplying a predetermined tap coefficient after delaying the signal, and an error signal obtained by subtracting the outputs of the common component extraction means and the difference component extraction means from the output signal of the microphone It is desirable to include a second coefficient update processing means for updating the value of the tap coefficient so that the power of the second is minimized. This makes it possible to reduce the processing amount of the second adaptive filter that performs the calculation related to the difference component.

  The first and second adaptive filters described above desirably perform processing in the time domain. When processing is performed in the time domain, the second adaptive filter corresponding to the difference component handles only the high-frequency component that converges quickly, so that it is possible to perform necessary calculations with a small number of taps. The processing of the multiplication means in which the tap coefficient is set and the coefficient update processing means for updating the value of the number of taps can be simplified.

  The first and second adaptive filters described above preferably include frequency component analysis means for analyzing the frequency component and perform processing in the frequency domain. When processing is performed in the frequency domain, the second adaptive filter corresponding to the difference component performs a necessary calculation only on the high frequency component obtained by the analysis of the frequency component. The processing of the set multiplication means and the coefficient update processing means for updating the value of the number of taps can be simplified.

[First Embodiment]
Hereinafter, an ASC (audio sound cancellation) apparatus according to a first embodiment to which the present invention is applied will be described with reference to the drawings. FIG. 1 is a diagram illustrating a detailed configuration of the ASC apparatus according to the first embodiment. The ASC device 100 shown in FIG. 1 cancels the audio component corresponding to the audio sound output from the left and right speakers 210 and 212 connected to the audio device 200 from the sound collected by the microphone 220. I do.

  The two speakers 210 and 212 are installed at symmetrical positions in the vehicle interior space. For example, a speaker 210 is installed on the left side of the rear part of the vehicle, and a speaker 212 is installed on the right side. The microphone 220 is installed at a position that is even from the left and right speakers 210 and 212 in the vehicle interior space and is close to the listener of the audio sound. When the driver or passenger in the passenger seat is considered as a listener, it may be installed on the ceiling corresponding to an intermediate position between the driver seat and the passenger seat, for example.

  In the present embodiment, the processing amount of the adaptive filter is reduced by paying attention to the symmetry of the acoustic transfer function. Specifically, since the speakers 210 and 212 are arranged in left and right symmetrical positions in the vehicle interior space, if the microphone 220 is installed at a position equidistant from the left and right speakers 210 and 212, the microphones 220 and 212 are arranged. The sound transfer function up to is almost symmetrical and is based on the idea that there are many parts that can be shared.

FIG. 2 is a block diagram showing a basic configuration of a conventional ASC apparatus corresponding to a 2-channel audio reproduction system. The ASC apparatus 100A shown in FIG. 2 uses an L channel reference signal x _L (n) to estimate an acoustic transfer function from the L channel speaker 210 to the microphone 220, a coefficient update processing unit 112, An FIR filter 120 for estimating an acoustic transfer function from the R channel speaker 212 to the microphone 220 using the channel reference signal x _R (n), and a coefficient update processing unit 122 are provided. The acoustic transfer function (first acoustic transfer function) from the L-channel speaker 210 to the microphone 220 is h _L (n), and the acoustic transfer function from the R-channel speaker 212 to the microphone 220 (second acoustic transfer function). ) Corresponds to h _R (n), the output signal of the microphone 220 corresponds to d (n), and the component corresponding to the L channel in d (n) corresponds to the R channel in d _L (n) and d (n). Let d _R (n) be the component to be
d (n) = d _L (n) + d _R (n)
^{_{= X L T (n) h}} L (n) + x R T (n) h R (n) ... (1)
It can be expressed as. As described above, the speakers 210 and 212 are arranged at symmetrical positions in the vehicle interior space, and when the microphone 220 is installed at an equal distance from the two speakers 210 and 212, the speakers 210 and 212 are respectively The acoustic transfer functions h _L (n) and h _R (n) up to the microphone 220 are substantially the same. Therefore, if the difference is Δh _RL (n),
h _R (n) = h _L (n) + (h _R (n) −h _L (n))
= H _L (n) + Δh _RL (n) (2)
It becomes. Substituting equation (2) into equation (1),
d (n) = x _L ^T (n) h _L (n) + x _R ^T (n) h _R (n)
^{_{= X L T (n) h}} L (n) + x R T (n) (h L (n) + Δh RL (n))
^{_{= (X L T (n)}} + x R T (n)) h L (n) + x R T (n) Δh RL (n)
... (3)
And can be transformed. When this equation is reflected in the time domain, the configuration block shown in FIG. 1 is obtained.

By the way, the left and right acoustic transfer functions h _L (n) and h _R (n) are not exactly the same even though they are substantially the same. Specifically, for low frequency components (for long wavelengths), it is considered to be almost the same, but the higher the frequency (the shorter the wavelength of the audio sound), the clearer the difference It is thought that it becomes. Thus, in the present embodiment, the difference Δh _RL (n) of the acoustic transfer function for the low frequency component of the audio sound is estimated to be negligible, and the difference Δh of the acoustic transfer function only for the high frequency component of the audio sound. _RL (n) is taken into account. Thus, when the band component to be processed is narrowed and only the high frequency component is focused, the number of taps of the FIR filter can be reduced.

Next, details of the ASC apparatus 100 shown in FIG. 1 will be described. As illustrated in FIG. 1, the ASC apparatus 100 includes a common component extraction unit 130, a difference component extraction unit 140, an addition unit 150, and a subtraction unit 160. The common component extraction unit 130 is for extracting a component corresponding to the first term on the right side of the equation (3), and has a configuration as an adaptive filter. Specifically, the common component extraction unit 130 receives an (N−1) -stage delay unit 10 to which an L channel reference signal x _L (n) is input and an R channel reference signal x _R (n). (N-1) stages of delay units 20 and two types of reference signals x _L (n) and x _R (n) input to and output from the (N-1) stages of delay units 10 and 20. ) And the like, and tap coefficients w _L (n), w _L (n−1),..., W _L (n−N + 1) for the outputs of the N adding units 30. ) Multiplication units 40 and (N−1) addition units 50 for adding the outputs of the N multiplication units 40.

By using the delay units 10 and 20 and the addition unit 30, the term (x _L ^T (n) + x _R ^T (n)) included in the first term on the right side of the equation (3) is calculated. Therefore, by adding by the adding unit 50 after passing through the multiplying unit 40 the output of the adder 30, the whole of the first term ^{_{(x L T (n) +}} x R T (n)) h L (n) Is calculated.

The difference extraction unit 140 is for extracting by focusing on the high frequency component of the component corresponding to the second term on the right side of the equation (3), and has a configuration as an adaptive filter. Specifically, the difference extraction unit 140 includes (M−1) stages of delay units 20 (M <N) to which an R channel reference signal x _R (n) is input, and (M−1) stages of these (M−1) stages. M multipliers 60 that operate as a high-pass filter as a whole to extract only high-frequency components included in the reference signal x _R (n) input / output to / from each stage of the delay unit 20, and M multipliers 60 Each of the M multipliers 70 that multiplies the respective outputs by tap coefficients w _R (n), w _R (n−1),..., W _R (n−M + 1), and M multipliers 70. And (M-1) adders 80 for adding outputs. The (M−1) stage delay units 20 described above use the first to (M−1) stages in the (N−1) stage delay units 20 included in the common component extraction unit 130. That is, (M−1) of the (N−1) delay units 20 are commonly used by the common component extraction unit 130 and the difference component extraction unit 140.

By passing through the delay unit 20, x _R ^T (n) included in the second term on the right side of the equation (3) is extracted. As described above, in this embodiment, by focusing only on the high frequency component, the number of taps of the FIR filter, that is, the number of delay units 20 is reduced from (N−1) to (M−1). As a result, the number of multiplication units 70 having tap coefficients w _R (n), w _R (n−1),..., W _R (n−M + 1) to be updated is also from (N−1). It can be reduced to (M-1).

  The above-described speakers 210 and 212 are the first and second speakers, the common component extraction unit 130 and the difference component extraction unit 140 are the component extraction unit, the common component extraction unit 130 is the common component extraction unit, and the difference component extraction unit 140. Corresponds to the difference component extraction means. Each component in the common component extraction unit 130 corresponds to the first adaptive filter, and each component in the difference component extraction unit 140 corresponds to the second adaptive filter. Furthermore, the multiplier 40 is the first multiplier, the coefficient update processor 42 is the first coefficient update processor, the multiplier 70 is the second multiplier, and the coefficient update processor 72 is the second coefficient update. Each corresponds to a processing means.

  As described above, the common component included in the audio sound of the L channel and the R channel included in the output signal d (n) of the microphone 220 is estimated using the common component extraction unit 130 and the difference component is extracted from the difference component extraction unit 140. So that the power of the error signal e (n) obtained by subtracting the result of adding these estimated values by the adding unit 150 from the actual output signal d (n) by the subtracting unit 160 is minimized. In addition, the tap coefficients of the common component extraction unit 130 and the difference component extraction unit 140 (multiplication values of the multiplication units 40 and 70) are updated by the coefficient update processing units 42 and 72, so that the output signal d ( It becomes possible to cancel the audio components of the L channel and the R channel included in n). In particular, by reducing the number of multiplication units 70 included in the difference component extraction unit 140, the processing amount of the coefficient update processing unit 72 that updates the value of the tap coefficient of each multiplication unit 70 can be reduced. Cost reduction can be realized by reducing the scale.

  FIG. 3 is a diagram illustrating a result of audio sound estimation using the ASC apparatus 100 of the present embodiment. In FIG. 3, the vertical axis represents the absolute value of the error signal e (n), and the horizontal axis represents the number of processing iterations (the number of times the multiplication value is updated by the coefficient update processing units 42 and 72). FIG. 4 is a diagram showing a result of audio sound estimation using the conventional ASC apparatus 100A shown in FIG. 2 for comparison. As apparent from FIGS. 3 and 4, even if the ASC device 100 of the present embodiment is used, it is possible to perform audio sound estimation equivalent to the ASC device 100A having the conventional configuration in almost the same processing time.

[Second Embodiment]
In the above-described embodiment, the ASC apparatus 100 that performs processing in the real time domain has been described. However, similar processing may be performed in the frequency domain. FIG. 5 is a diagram illustrating a configuration of an ASC apparatus according to the second embodiment that performs processing in the frequency domain. In addition, the same code | symbol is attached | subjected about what performs the operation | movement fundamentally the same or similar to the structure shown in FIG. 1, and detailed description shall be abbreviate | omitted.

The ASC device 100B illustrated in FIG. 5 includes a common component extraction unit 130A, a difference component extraction unit 140A, an addition unit 150, and a subtraction unit 160. The common component extraction unit 130A is for extracting a component corresponding to the first term on the right side of the equation (3). The basic configuration of having a configuration as an adaptive filter is the common component extraction unit 130 shown in FIG. Is the same. The common component extraction unit 130A includes (N−1) stages of delay units 10 to which an L-channel reference signal x _L (n) is input, and reference signals x _L (input / output to / from each stage of the delay unit 10). n), x _L (n−1),..., x _L (n−N + 1) to perform a discrete Fourier transform (DFT) to analyze the frequency component of x _L (n), The (N-1) stage delay units 20 to which the R channel reference signal x _R (n) is input, and the reference signals x _R (n) and x _R (n−) input / output to / from each stage of the delay unit 20. 1),..., X _R (n−N + 1) is used to perform a discrete Fourier transform to analyze the frequency component of x _R (n), and the analysis of these two DFT processing units 12 and 22 Multiple additions that add together the low-frequency components obtained by 30, low tap coefficient for each frequency for the high frequency component using the output of the adder 30 components using high-frequency components obtained by the analysis of the DFT processing unit _{12 W L (0, n)} , W _L (1, n),..., W _L (N−2, n), N _L multiplying units 40 that multiply W _L (N−1, n), and outputs of the N multiplying units 40 And (N-1) adders 50 to be added.

By using the delay units 10 and 20, the DFT processing units 12 and 22, and the addition unit 30, the term (x _L ^T (n) + x _R ^T (n)) included in the first term on the right side of the equation (3) The calculation is performed in the frequency domain (for each frequency component). Therefore, by adding by the adding unit 50 after passing through the multiplying unit 40 the output of the adder 30, the whole of the first term ^{_{(x L T (n) +}} x R T (n)) h L (n) Is calculated.

The difference component extraction unit 140A is for extracting a component corresponding to the second term on the right side of the equation (3). The basic configuration of having a configuration as an adaptive filter is the common component extraction unit 130 shown in FIG. Is the same. Specifically, the difference component extraction unit 140A includes (N−1) stages of delay units 20 to which the R channel reference signal x _R (n) is input, and these (N−1) stages of delay units 20. Frequency components of x _R (n) by performing discrete Fourier transform using reference signals x _R (n), x _R (n−1),..., X _R (n−N + 1) input / output to / from each stage. DFT processing unit 22 for analyzing the frequency, and a plurality of high frequency components obtained by the analysis of DFT processing unit 22 are multiplied by tap coefficients W _R (N−2, n) and W _R (N−1, n). Multiplication unit 70, and an addition unit 80 for adding the outputs of the multiplication unit 70. The (N−1) -stage delay unit 20 and the DFT processing unit 22 described above commonly use the (N−1) -stage delay unit 20 and the DFT processing unit 22 included in the common component extraction unit 130A. The above-described DFT processing units 12 and 22 correspond to frequency component analysis means.

  In this way, by calculating the difference component by paying attention to the high frequency component obtained by performing the frequency analysis using the DFT processing unit 22, the number of multiplication units 70 included in the difference component extraction unit 140A can be reduced. It is possible to reduce the amount of processing of the coefficient update processing unit 72 that updates the value of the tap coefficient of each multiplication unit 70, so that the hardware scale can be reduced and cost reduction can be realized.

  In addition, this invention is not limited to the said embodiment, A various deformation | transformation implementation is possible within the range of the summary of this invention. For example, in the above-described embodiment, the case where the speakers 210 and 212 and the microphone 220 are installed in the vehicle interior space has been described, but the case where the speaker or the like is installed in a space other than the vehicle interior space, for example, indoors or outdoors. Similarly, the present invention can be applied.

  In the above-described embodiment, the case where the audio component included in the output signal of the microphone 220 is canceled has been described. However, the present invention estimates the spatial transfer function by paying attention to the symmetry of the arrangement of the speaker and the microphone. Therefore, the signal to be processed may be a sound other than the audio sound or other signals.

It is a figure which shows the detailed structure of the ASC apparatus of 1st Embodiment. It is a block diagram which shows the basic composition of the conventional ASC apparatus corresponding to a 2-channel audio reproduction system. It is a figure which shows the result of having performed the estimation of the audio sound using the ASC apparatus of this embodiment. It is a figure which shows the result of having estimated the audio sound using the ASC apparatus of the conventional structure shown in FIG. 2 for the comparison. It is a figure which shows the structure of the ASC apparatus of 2nd Embodiment which processes in a frequency domain.

Explanation of symbols

10, 20 Delay unit 12, 22 DFT (Discrete Fourier Transform) processing unit 30, 50, 80 Addition unit 40, 60, 70 Multiplication unit 42, 72 Coefficient update processing unit 100, 100A, 100B ASC (Audio Sound Cancellation) device 130, 130A Common component extraction unit 140, 140A Difference component extraction unit 150 Addition unit 160 Subtraction unit 200 Audio device 210, 212 Speaker 220 Microphone

Claims (5)

First and second speakers installed at symmetrical positions in the acoustic space;
A microphone disposed equidistant from the first and second speakers;
By estimating a first acoustic transfer function from the first speaker to the microphone and a second acoustic transfer function from the second speaker to the microphone, the first acoustic transfer function is included in the output signal of the microphone. And a component extracting means for extracting a component corresponding to the output sound of the second speaker,
The component extraction means performs extraction corresponding to a common component of the first and second acoustic transfer functions, and performs extraction corresponding to a difference component of the first and second acoustic transfer functions. A sound component characterized by comprising difference component extraction means for paying attention to a band component. In claim 1,
The common component extraction means includes a first adaptive filter that estimates an acoustic transfer function corresponding to the common component using a signal input to each of the first and second speakers as a reference signal,
The difference component extraction unit includes a second adaptive filter that estimates an acoustic transfer function corresponding to the common component using a signal input to the second speaker as a reference signal,
An acoustic apparatus, wherein the number of taps of the second adaptive filter is set to be smaller than the number of taps of the first adaptive filter. In claim 2,
The first adaptive filter includes N first multiplication means for multiplying a predetermined tap coefficient after delaying an input reference signal, the common component extraction means and the difference component from an output signal of the microphone. First coefficient update processing means for updating the value of the tap coefficient so that the power of the error signal obtained by subtracting the outputs of the extraction means is minimized,
The second adaptive filter includes M second multiplication means for multiplying a predetermined tap coefficient after delaying an input reference signal, and extracting the common component from the output signal of the microphone. And a second coefficient update processing means for updating the value of the tap coefficient so that the power of the error signal obtained by subtracting the outputs of the difference component extraction means is minimized. In claim 2,
The acoustic device according to claim 1, wherein the first and second adaptive filters perform processing in a time domain. In claim 2,
The first and second adaptive filters include frequency component analysis means for analyzing frequency components, and perform processing in the frequency domain.

Images