JP5725088B2 - Sound collection device and sound emission collection system - Google Patents

Description

  The present invention relates to a sound collection device and a sound emission and collection system that collect sound in a plurality of directions.

  Conventionally, various sound collection devices for collecting a speaker's voice have been proposed (for example, see Patent Document 1).

  For example, the sound collection device of Patent Document 1 includes a microphone array including a plurality of microphones, performs a delay process on the audio signal collected by each microphone, and sets different directions as the directivity axial directions. A plurality of sound collecting beam signals having sound directivity are generated. Then, the sound collection device selects the sound collection beam signal having the highest signal level from among the plurality of sound collection beam signals, removes the echo from the selected sound collection beam signal, and transmits it to the communication partner. The main speaker's voice is transmitted to the other party.

Japanese Patent Laid-Open No. 2002-238091

  However, the collected sound beam signal includes a speaker's voice and a return sound based on the voice emitted from the own apparatus. For this reason, the sound collecting device cannot accurately select the main speaker's voice when the return sound is larger than the speaker's voice.

  In such a case, the sound collection device selects the sound of the main speaker accurately if the sound collection beam signal to be transmitted is selected based on the signal level of each sound collection beam signal after removing the echo. be able to. However, the processing for removing the echoes of all collected beam signals (for example, 6 directions) has a heavy load, and it is impossible to practically remove the omnidirectional echoes.

  Therefore, an object of the present invention is to provide a sound collection device and a sound emission and collection system that can accurately select the voice of a main speaker without applying a processing load.

  The sound collection device of the present invention generates a sound collection signal for each of a plurality of different directions and estimates the direction of a sound source (for example, a main speaker). The sound collection device includes a first echo cancellation unit and a plurality of second echo cancellation units. The first echo canceling means removes echoes of the collected sound signal from the estimated direction. The second echo canceling unit has a simpler configuration than the first echo canceling unit, and removes the echo from the collected sound signal for each direction. The sound collection device estimates the direction of the sound source based on the signal level of the sound collection signal after processing by the plurality of second echo cancellation means. Note that the echo canceling means having a simple configuration is one that removes echoes from the collected sound signal after downsampling, or one that removes echoes using an adaptive filter with a small number of taps.

  Thereby, since the sound collection device estimates the direction of the sound source using the sound collection signal after the echo is simply removed, it is possible to accurately estimate the direction of the sound source without applying a processing load.

  The first echo canceling means and the plurality of second echo canceling means of the sound collecting device of the present invention each include an adaptive filter and a filter coefficient estimating unit that estimates a filter coefficient of the adaptive filter. The filter coefficient estimation means of the first echo cancellation means performs a process of updating the filter coefficient using the filter coefficient of the second echo cancellation means from which the echo of the collected sound signal from the estimated sound source direction is removed as an initial value.

  As a result, the sound collection device can remove the echo of the collected sound signal from the sound source direction using the filter coefficient used in the estimation of the sound source direction as an initial value, thereby reducing the estimation time of the filter coefficient. Can do. Therefore, the sound collection device can remove the echo from the initial state.

  Furthermore, the first echo cancellation means of the sound collecting device of the present invention stores the filter coefficient of the adaptive filter for each direction. The filter coefficient estimating unit of the first echo canceling unit performs the initial process using the filter coefficient of the second echo canceling unit as an initial value as described above only when no filter coefficient is stored.

  As a result, the first echo canceling means of the sound collecting device sets the filter coefficient used at the time of estimating the sound source direction as the initial value only when the filter coefficient is not stored in the storage unit, and otherwise the previously used filter coefficient Since the coefficients are set as initial values, echoes can be immediately removed even if the environment (speaker) changes.

  The present invention is not limited to the sound collection device, and may be a sound emission / collection system including a sound emission / collection device including a communication unit and a sound emission unit. The communication means transmits the sound collection signal from which the echo is removed by the first echo cancellation means to another sound emission and collection device. The second echo canceling unit may generate a pseudo regression signal based on the downsampled sound emission signal.

  The sound collection device and the sound emission and collection system of the present invention can accurately estimate the direction of a sound source (for example, a main speaker) without applying a processing load.

It is a block diagram which shows the function and structure of a sound emission and collection apparatus. It is explanatory drawing for demonstrating the sound collection direction for every sound collection beam signal. It is a block diagram which shows the function and structure of an echo cancellation part. It is a figure which shows an example of the filter coefficient for every sound collection direction of an adaptive filter. It is a block diagram which shows the function and structure of the sound emission and collection apparatus which concern on other embodiment.

  A sound emitting and collecting apparatus 1 according to an embodiment of the present invention will be described with reference to FIGS. The sound emission and collection device 1 is connected to another sound emission and collection device via a network or the like. The sound emission and collection device 1 receives an audio signal from another sound emission and collection device as a sound emission signal and emits the sound from the speaker SP. The sound emission and collection device 1 collects sound with the microphones MIC1 to MIC3 and generates sound collection beam signals from a plurality of directions. Then, the sound emission and collection device transmits a sound collection beam signal from the main speaker's direction to another sound emission and collection device.

  First, the function and configuration of the sound emission and collection device 1 will be described with reference to FIGS. FIG. 1 is a block diagram showing the function and configuration of the sound emission and collection device. FIG. 2 is an explanatory diagram for explaining a sound collection direction for each sound collection beam signal. The sound emission and collection device 1 includes a speaker SP, microphones MIC1 to MIC3, a communication control unit 11, a sound collection control unit 12, a downsampling unit (hereinafter referred to as a DS unit) 13, and a downsampling unit (hereinafter referred to as a DS unit). .) 14A to 14C, echo canceling unit (corresponding to the second echo canceling unit of the present invention) 15A to 15C, azimuth estimating unit 16, control unit 17, collected sound signal selecting unit 18, and echo canceling unit (the present invention) This corresponds to the first echo canceling unit.) 19.

  The communication control unit 11 is connected to another sound emitting and collecting device via a network, and performs control related to communication with the other sound emitting and collecting device. Specifically, the communication control unit 11 receives the sound emission signal FE from another sound emission and collection device, and outputs the sound emission signal FE to the DS unit 13 and the speaker SP described later via the echo cancellation unit 19. The speaker SP emits sound based on the sound emission signal FE. Further, the communication control unit 11 transmits a sound collection beam signal NE1 'input from an echo cancellation unit 19 described later to another sound emission and collection device.

  The microphones MIC <b> 1 to MIC <b> 3 collect surrounding sounds, generate sound collection signals, and output them to the sound collection control unit 12. The number of microphones is not limited to three.

  The sound collection control unit 12 performs a delay process on the sound collection signals from the microphones MIC1 to MIC3, and has a plurality of sound collection beams having different directions as the central direction of the sound collection directivity as shown in FIG. Signals NE1 to NE3 are generated. Hereinafter, the sound collection directions of the sound collection beam signals NE1 to NE3 are D1 to D3.

  Then, the sound collection control unit 12 outputs the sound collection beam signals NE1 to NE3 to the DS units 14A to 14C, respectively, and outputs the sound collection beam signals NE1 to NE3 to the sound collection signal selection unit 18. The number of sound collection beam signals generated by the sound collection control unit 12 is not limited to three. The sound collection control unit 12 is not an essential configuration. In this case, each of the microphones MIC <b> 1 to MIC <b> 3 collects sound from different directions, generates a sound collection signal, and outputs the sound collection signal to the DS units 14 </ b> A to 14 </ b> C and the sound collection signal selection unit 18.

  The DS units 14A to 14C include low-pass filters, perform downsampling of the input sound collection beam signals NE1 to NE3, and output the downsampled sound collection beam signals DNE1 to DNE3 to the echo cancellation units 15A to 15C, respectively. To do. For example, the DS units 14A to 14C down-sample the collected sound beam signals DNE1 to NE3 sampled at a sampling frequency of 20 kHz into signals having a sampling frequency of 10 kHz.

  The DS unit 13 includes a low-pass filter, performs downsampling of the input sound emission signal FE, and outputs the sound emission signal DFE after downsampling to the echo cancellation units 15A to 15C. For example, the DS unit 13 down-samples the sound emission signal FE sampled at a sampling frequency of 20 kHz into a signal having a sampling frequency of 10 kHz.

  The echo cancellation units 15A to 15C generate a pseudo regression sound signal that is a pseudo signal of a wraparound component from the speaker SP to each of the microphones MIC1 to MIC3, based on the sound output signal DFE after downsampling. The echo cancellation units 15A to 15C remove echoes by subtracting the pseudo regression sound signals from the down-sampled sound collection beam signals DNE1 to DNE3, respectively. Then, the echo cancellation units 15A to 15C output the collected sound beam signals DNE1 'to DNE3' after the echo is removed to the azimuth estimation unit 16.

  Since the echo canceling units 15A to 15C are for removing echoes of the collected sound beam signals DNE1 to DNE3 after downsampling, the echo canceling units 15A to 15C have a simpler configuration than the echo canceling unit 19 and do not apply a processing load. Echo can be removed. Further, the echo cancellation units 15 </ b> A to 15 </ b> C may have fewer taps than the echo cancellation unit 19. The detailed functions and configurations of the echo cancellation units 15A to 15C will be described later.

  The azimuth estimating unit 16 selects a sound collecting beam signal having the highest signal level from the sound collecting beam signals DNE1 'to DNE3' after echo cancellation. In the following description, it is assumed that the direction estimation unit 16 has selected the sound collection beam signal DNE1 'from the sound collection direction D1. Then, the direction estimation unit 16 acquires the sound collection direction D1 of the selected sound collection beam signal DNE1 'and outputs it to the control unit 17.

  The control unit 17 controls the sound collection signal selection unit 18 and the echo cancellation unit 19 based on the sound collection direction D1 input from the direction estimation unit 16.

  The sound collection signal selection unit 18 selects the sound collection beam signal NE1 from the sound collection direction D1 input from the control unit 17 from the sound collection beam signals NE1 to NE3 input from the sound collection control unit 12. And output to the echo canceling unit 19.

  Based on the sound emission signal FE, the echo cancellation unit 19 generates a pseudo regression sound signal that is a pseudo signal of a wraparound component from the speaker SP to each of the microphones MIC1 to MIC3, and is input from the sound collection signal selection unit 18 The echo is removed by subtracting the pseudo regression signal from the collected sound beam signal NE1. Then, the echo cancellation unit 19 outputs the collected sound beam signal NE1 'after the echo is removed to the communication control unit 11. The detailed function and configuration of the echo cancellation unit 19 will be described later.

  As described above, since the echo cancellation units 15A to 15C remove echoes from the collected sound beam signals DNE1 to DNE3 after downsampling, the processing load can be reduced. Further, the sound emitting and collecting apparatus 1 acquires the direction of the main speaker using the collected sound beam signals DNE1 ′ to DNE3 ′ after removing the echo, and collects the sound beam from the main speaker's direction. Since the signal is selected, the voice of the main speaker can be accurately selected. Therefore, the sound emission and collection device 1 can accurately select the voice of the main speaker without applying a processing load.

  Next, details of functions and configurations of the echo cancellation units 15A to 15C and the echo cancellation unit 19 will be described with reference to FIGS. FIG. 3 is a block diagram showing the function and configuration of the echo cancellation unit. FIG. 4 is a diagram illustrating an example of a filter coefficient for each sound collection direction of the adaptive filter. The echo cancellation units 15A to 15C and the echo cancellation unit 19 have the same function and configuration. Hereinafter, the echo canceling unit 19 will be described as an example.

  As shown in FIG. 3, the echo cancellation unit 19 includes a storage unit 21, a filter coefficient estimation unit 22, an adaptive filter 23, and an addition unit 24.

  The storage unit 21 temporarily stores a coefficient list 211 as shown in FIG. The coefficient list 211 describes filter coefficients for each sound collection direction, and is referred to by the filter coefficient estimation unit 22. Note that the filter coefficients stored in the coefficient list 211 are reset when the power of the sound emission and collection device 1 is turned on / off.

  The filter coefficient estimation unit 22 estimates a transfer function of the acoustic transmission system (acoustic propagation path from the speaker SP to each of the microphones MIC1 to MIC3), and sets the filter coefficient of the FIR filter with the estimated transfer function. At this time, the filter coefficient estimation unit 22 acquires a filter coefficient corresponding to the sound collection direction D1 input from the control unit 17 from the coefficient list 211, and calculates the filter coefficient using the filter coefficient as an initial value. Further, the filter coefficient estimating unit 22 updates the filter coefficient using an adaptive algorithm based on the collected sound beam signal NE1 'and the sound emission signal FE output from the adder unit 24. Then, the filter coefficient estimation unit 22 outputs the calculated filter coefficient to the adaptive filter 23.

  The adaptive filter 23 includes a digital filter such as an FIR filter, and generates a pseudo regression sound signal using the filter coefficient input from the filter coefficient estimation unit 22. The adaptive filter 23 outputs the generated pseudo regression sound signal to the adding unit 24.

  The adding unit 24 outputs a sound collection beam signal NE1 'obtained by subtracting the pseudo regression sound signal input from the adaptive filter 23 from the sound collection beam signal NE1.

  The storage unit 21 is not an essential configuration. However, since the echo cancellation unit 19 needs to change the initial value of the filter coefficient when the sound collection direction is switched, it is preferable to include the storage unit 21.

  Further, the filter coefficient estimating unit 22 of the echo canceling units 15A to 15C performs an adaptive algorithm based on the collected sound beam signals DNE1 ′ to DNE3 ′ output from the adding unit 24 and the sound emission signal DFE after downsampling, respectively. To update the filter coefficients.

  As described above, since the echo cancellation unit 19 stores the filter coefficient for each sound collection direction in the storage unit 21, the environment (speaker) changes when the sound collection direction is switched. The filter coefficient is acquired from the storage unit 21, and the adaptive filter is estimated. For this reason, the echo canceling unit 19 can shorten the estimation of the adaptive filter, and can immediately remove the echo even if the environment (speaker) changes.

  The echo cancellation unit 19 may perform echo cancellation using the filter coefficients acquired from the echo cancellation units 15A to 15C. In this case, the filter coefficients of the echo cancellation units 15A to 15C are upsampled and used. FIG. 5 is a block diagram illustrating functions and configurations of a sound emission and collection device according to another embodiment. As shown in FIG. 5, the control unit 17 refers to the coefficient list 211 in the storage unit 21 of the echo cancellation unit 19 when the sound emission and collection device 1 is turned on or when the sound collection direction is switched. Only when the filter coefficient of the sound collection direction input from the direction estimation unit 16 is not described in the coefficient list 211, the control unit 17 removes the echo of the sound collection beam signal from the sound collection direction. The filter coefficients are acquired from the units 15 </ b> A to 15 </ b> C and output to the filter coefficient estimation unit 22 of the echo cancellation unit 19. Then, the filter coefficient estimation unit 22 calculates a filter coefficient using the filter coefficient input from the control unit 17 as an initial value. As a result, the echo cancel unit 19 calculates the filter coefficient using the filter coefficient when the echo of the downsampled sound collecting beam signals DNE1 to DNE3 is removed as an initial value, and therefore shortens the estimation time of the filter coefficient. Can do. For this reason, the echo cancellation part 19 can remove an echo from an initial state.

DESCRIPTION OF SYMBOLS 1 ... Sound emission and collection device, 11 ... Communication control part, 12 ... Sound collection control part, 13, 14 ... DS part, 15, 19 ... Echo cancellation part, 16 ... Direction estimation part, 17 ... Control part, 18 ... Sound collection Signal selection unit, 21 ... storage unit, 211 ... coefficient list, 22 ... filter coefficient estimation unit, 23 ... adaptive filter, 24 ... addition unit, MIC1 to MIC3 ... microphone, SP ... speaker

Claims (7)

A sound collection device comprising echo cancellation means for removing echoes of a sound collection signal based on a sound emission signal,
A plurality of sound collecting means for collecting sound of a plurality of directions and generating a sound collection signal for each direction;
Direction estimation means for estimating the direction of the sound source based on the signal level of the collected sound signal for each direction;
First echo cancellation means for removing echoes of the collected sound signal from the direction of the sound source estimated by the direction estimation means;
Echo is removed from the collected sound signal collected by the sound collecting means,
A plurality of second echo canceling means having a simpler configuration than the first echo canceling means;
With
The direction estimating means estimates the direction of a sound source based on the signal level of the collected sound signal after processing by the plurality of second echo canceling means. Each of the first echo cancellation means and the plurality of second echo cancellation means includes an adaptive filter and a filter coefficient estimation means for estimating a filter coefficient of the adaptive filter,
The second echo cancellation means has a smaller number of filter coefficients than the first echo cancellation means,
The sound collection device according to claim 1. Downsampling means for downsampling the collected sound signal picked up by the sound collecting means;
The second echo canceling means removes an echo from the downsampled sound pickup signal;
The sound collection device according to claim 1. The filter coefficient estimation means of the first echo cancellation means is
An initial process is performed in which the filter coefficient of the adaptive filter of the second echo cancellation unit that removes the echo of the collected sound signal of the direction estimated by the direction estimation unit is an initial value.
The sound collection device according to any one of claims 1 to 3. The first echo cancellation means further comprises storage means for storing the filter coefficient of the adaptive filter for each direction,
The filter coefficient estimation means of the first echo cancellation means performs the initial process only when no filter coefficient is stored in the storage means.
The sound collection device according to claim 4. The sound collecting device according to any one of claims 1 to 5 constitutes a sound emitting and collecting device including a communication unit and a sound emitting unit,
A sound emission and collection system comprising the sound emission and collection device,
The communication means transmits the sound collection signal from which echo is removed by the first echo cancellation means to another sound emission and collection device.
Sound emission and collection system. The second echo cancellation means generates a pseudo regression signal based on the downsampled sound emission signal;
The sound emission and collection system according to claim 6.

Images