Interference sound suppression device, interference sound suppression method, and interference sound suppression program
By inserting a plurality of 0s into the input signal instead of inserting a plurality of 0s into the reference signal, the problem of difficulty in taking into account both the call quality and the calculation amount in the prior art is solved, and the effect of improving the call quality and reducing the calculation amount is achieved.
Patent Information
- Application Number
- CN202280094290.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-03-29
- Filing Date
- 2022-12-28
- Publication Date
- 2024-11-12
AI Technical Summary
The prior art is difficult to take into account the improvement of call quality and the reduction of computational amount. Especially when multiple zeros are inserted during frequency domain transformation to reduce the influence of aliasing distortion, the frequency resolution and computational amount of frequency domain transformation will increase.
By inserting multiple 0s into the input signal instead of multiple 0s in the reference signal, the data amount of the input signal is increased to suppress the interference component, and the effects of aliasing distortion are reduced through frequency domain transformation and suppression coefficient calculations, ensuring frequency resolution rate consistency.
Improve call quality and reduce computational volume, reduce the impact of aliasing distortion while maintaining sufficient frequency resolution.
Smart Images

Figure CN118947137A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to a technique for suppressing interfering sounds included in an input signal obtained from a microphone. Background Art
[0002] For example, the noise suppression device shown in Patent Document 1 stores the input signal by a given number of samples N each time. Whenever the number of the latest samples among the stored samples reaches a predetermined number of samples M (<N), a transform frame for transforming the input signal into frequency domain coefficients is generated, and each sample data of the transform frame is transformed into the frequency domain. And, when transforming into the frequency domain, the noise suppression device only performs arithmetic processing related to the updated M samples, and uses the arithmetic results of the previous frame for other data by copying. And, the noise suppression device suppresses a noise signal from the above input signal in the frequency domain, transforms the noise-suppressed noise-suppression completed signal into the time domain, generates an addition frame, which is used to generate an output signal, overlaps and adds the M samples of the addition frame and the M samples of the addition frame one frame before the addition frame, and generates an output signal.
[0003] In addition, for example, the echo cancellation device shown in Patent Document 2 analyzes the frequency components of the sum of the reproduction signals and the frequency components of the pickup signal, groups the respective frequency components, and estimates the amplitude spectrum of the echo of the group according to the amplitude ratio for each group. And, the echo cancellation device calculates the frequency components of the echo cancellation signal according to the amplitude ratio of each frequency component of the echo amplitude spectrum estimated to be the frequency components of the pickup signal, transforms the frequency components of the echo cancellation signal into the time domain and outputs them.
[0004] However, in the above prior art, it is difficult to balance the improvement of call quality and the reduction of the amount of computation, and further improvement is required.
[0005] Prior Art Documents
[0006] Patent Documents
[0007] Patent Document 1: Japanese Patent No. 4253232 Gazette
[0008] Patent Document 2: Japanese Patent No. 4478045 Gazette Summary of the Invention
[0009] The present disclosure is made to solve the above problems, and an object thereof is to provide a technique capable of improving call quality and reducing the amount of computation.
[0010] The interference sound suppression device involved in the present disclosure comprises: an input signal acquisition unit, which acquires an input signal obtained from a microphone; a reference signal acquisition unit, which acquires a reference signal; a data insertion unit, which inserts a plurality of zeros into the acquired input signal, thereby increasing the data volume of the input signal compared with the data volume of the reference signal; an interference sound suppression unit, which suppresses the interference sound component contained in the input signal using the input signal with the increased data volume and the acquired reference signal; a data deletion unit, which restores the data volume of the input signal by deleting the inserted plurality of zeros from the input signal with the suppressed interference sound component; and an output unit, which outputs the input signal with the restored data volume.
[0011] According to the present disclosure, call quality can be improved and the amount of calculation can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS
[0012] Figure 1 This is a diagram showing the structure of a communication device in Embodiment 1 of the present disclosure.
[0013] Figure 2 It is a diagram showing the structure of the echo suppression device in Embodiment 1 of the present disclosure.
[0014] Figure 3 This is a flowchart for explaining the operation of the echo suppression device in Embodiment 1 of the present disclosure.
[0015] Figure 4 This is a diagram showing the structure of a sound processing device in Embodiment 2 of the present disclosure.
[0016] Figure 5 It is a diagram showing the structure of a noise suppression device in Embodiment 2 of the present disclosure. DETAILED DESCRIPTION
[0017] (Knowledge that forms the basis of this disclosure)
[0018] In the past, there was a sound-reinforced two-way communication system using a microphone and a speaker. In such a sound-reinforced two-way communication system, the sound spoken by the speaker on the transmitting side is input into the microphone on the transmitting side, sent as a transmitting signal to the device on the receiving side via the communication line, and reproduced by the speaker on the receiving side. The sound reproduced by the speaker on the receiving side propagates in the space on the receiving side and is input into the microphone on the receiving side, and is sent to the transmitting side. At this time, the sound of the speaker's own speech that has passed through the communication line and propagated in the space on the receiving side is reproduced from the speaker on the transmitting side. In this way, the sound propagated between the microphones from the speaker on the receiving side is called acoustic echo, which leads to deterioration of call quality.
[0019] Therefore, in a two-way communication system of a sound reinforcement type, an echo suppressor for suppressing acoustic echo is used. In the echo suppressor, filter processing for suppressing acoustic echo components is performed on input data in the frequency domain using discrete Fourier transform (DFT). However, there is a concern that due to the filter processing in the frequency domain, a phenomenon called aliasing distortion of the second half of the input data is generated, which is aliased to the first half. The generation of aliasing distortion leads to deterioration of call quality. Therefore, in order to reduce the influence of aliasing distortion, the following operation is performed: by inserting multiple 0s of the same data amount as the input data into the input data (0 padding), the data length of the input data is increased to twice. As a result, multiple 0 data are aliased as distortion, and the influence of aliasing distortion on the input data is reduced.
[0020] However, when multiple zeros are inserted into the input data, the call quality can be prevented from being degraded. However, since the data length of the input data processed by the filter processing is doubled, the amount of calculation is also doubled, and the increase in the amount of calculation becomes a problem. In particular, the input data includes not only the input signal obtained from the microphone, but also a reference signal for suppressing interference sounds included in the input signal. Therefore, the data length of each of the input signal and the reference signal is doubled.
[0021] In the above-mentioned Patent Document 1, when the frequency domain is transformed, only the calculation processing related to the updated M samples is performed, and the calculation results of the previous frame are copied and used for other data, thereby reducing the amount of calculation. In addition, in the above-mentioned Patent Document 2, only the amplitude ratio between the frequency component of the sound pickup signal and the estimated echo amplitude spectrum is used to calculate the frequency component of the echo cancellation signal, thereby reducing the amount of calculation.
[0022] However, in Patent Documents 1 and 2, there is a concern that when a plurality of zeros are inserted during frequency domain transformation in order to reduce the influence of aliasing distortion, the frequency resolution of the frequency domain transformation will be doubled, and the amount of calculation will also be doubled. Therefore, in the conventional technology, it is difficult to achieve both improvement of call quality and reduction of calculation.
[0023] In order to solve the above problems, the following technology is disclosed.
[0024] (1) An interference sound suppression device according to one embodiment of the present invention comprises: an input signal acquisition unit for acquiring an input signal obtained from a microphone; a reference signal acquisition unit for acquiring a reference signal; a data insertion unit for inserting a plurality of zeros into the acquired input signal, thereby increasing the data volume of the input signal compared with the data volume of the reference signal; an interference sound suppression unit for suppressing an interference sound component included in the input signal using the input signal with the increased data volume and the acquired reference signal; a data deletion unit for restoring the data volume of the input signal by deleting the inserted plurality of zeros from the input signal with the suppressed interference sound component; and an output unit for outputting the input signal with the restored data volume.
[0025] According to this structure, by inserting a plurality of zeros into the input signal, the influence of aliasing distortion generated when suppressing the interference sound component included in the input signal can be reduced, thereby improving the call quality. In addition, by inserting a plurality of zeros only into the input signal and not inserting a plurality of zeros into the reference signal, the amount of calculation can be reduced compared to the case where a plurality of zeros are inserted into both the input signal and the reference signal.
[0026] (2) In the interference sound suppression device described in (1) above, the data inserting unit may increase the number of samples of the input signal to twice or more the number of samples of the reference signal.
[0027] According to this configuration, since the number of samples of the input signal increases to more than twice the number of samples of the reference signal, sufficient frequency resolution can be ensured and the influence of aliasing distortion can be reliably reduced.
[0028] (3) In the interference sound suppression device described in (1) or (2) above, the reference signal may be a reproduced signal output to a speaker.
[0029] According to this configuration, since the reference signal is a reproduced signal output to the speaker, it is possible to suppress the acoustic echo component included in the input signal.
[0030] (4) In the interference sound suppression device described in (1) or (2) above, the reference signal may be a noise signal obtained from a microphone different from the microphone.
[0031] According to this configuration, since the reference signal is a noise signal obtained from a microphone different from the microphone, it is possible to suppress the noise component included in the input signal.
[0032] (5) In the interference sound suppression device described in (1) or (2) above, the reference signal may be a noise signal obtained from the microphone from a given direction.
[0033] According to this configuration, since the reference signal is a noise signal obtained from a microphone and coming from a given direction, it is possible to suppress the noise component included in the input signal.
[0034] (6) In the interference sound suppression device described in any one of (1) to (5) above, the input signal acquisition unit may acquire the input signal in the time domain, the reference signal acquisition unit may acquire the reference signal in the time domain, and the interference sound suppression unit may include the following units: a first frequency domain transformation unit, which transforms the input signal in the time domain with an increased data volume into an input signal in the frequency domain; a second frequency domain transformation unit, which transforms the acquired reference signal in the time domain into a reference signal in the frequency domain; and a sample number reduction unit, which reduces the number of samples of the transformed input signal in the frequency domain to the same number as the reference signal. a coefficient calculation unit, which calculates a suppression coefficient for suppressing the interference sound component based on the input signal in the frequency domain whose sample number is reduced and the reference signal in the frequency domain after transformation; a sample number increasing unit, which increases the number of samples of the calculated suppression coefficient by the same number of samples as the input signal whose data amount is increased; a multiplication unit, which multiplies the input signal in the frequency domain after transformation by the suppression coefficient whose sample number is increased; and a time domain transformation unit, which transforms the input signal in the frequency domain multiplied by the suppression coefficient into an input signal in the time domain.
[0035] According to this structure, since the number of samples of the input signal in the frequency domain is reduced to the same number of samples as the reference signal, the frequency resolution of the input signal can be made consistent with the frequency resolution of the reference signal when calculating the suppression coefficient. In addition, since the number of samples of the calculated suppression coefficient is increased to the same number of samples as the input signal with an increased data amount, the frequency resolution of the suppression coefficient can be made consistent with the frequency resolution of the input signal.
[0036] (7) In the interference sound suppression device described in any one of (1) to (6) above, the reference signal acquisition unit may also acquire multiple reference signals, and the interference sound suppression unit uses the input signal with an increased data volume and the acquired multiple reference signals to suppress the interference sound component contained in the input signal.
[0037] According to this structure, in a multi-channel call system using multiple speakers, since multiple zeros are inserted only in the input signal and multiple zeros are not inserted in multiple reference signals, the amount of calculation can be greatly reduced compared to the case where multiple zeros are inserted in all input signals and multiple reference signals.
[0038] In addition, the present disclosure can be implemented not only as an interference sound suppression device having the characteristic structure as described above, but also as an interference sound suppression method that performs characteristic processing corresponding to the characteristic structure of the interference sound suppression device. In addition, it can also be implemented as a computer program that causes a computer to execute the characteristic processing included in such an interference sound suppression method. Therefore, the following another method can also achieve the same effect as the above-mentioned interference sound suppression device.
[0039] (8) Another embodiment of the present invention relates to an interference sound suppression method of a computer, and the interference sound suppression method includes the following operations: obtaining an input signal obtained from a microphone, obtaining a reference signal, inserting a plurality of zeros into the obtained input signal so that the data volume of the input signal is increased compared with the data volume of the reference signal, suppressing the interference sound component contained in the input signal by using the input signal with the increased data volume and the obtained reference signal, restoring the data volume of the input signal by deleting the inserted plurality of zeros from the input signal in which the interference sound component is suppressed, and outputting the input signal with the restored data volume.
[0040] (9) Another embodiment of the present invention relates to an interference sound suppression program that enables a computer to function as the following units: an input signal acquisition unit that acquires an input signal obtained from a microphone; a reference signal acquisition unit that acquires a reference signal; a data insertion unit that inserts a plurality of zeros into the acquired input signal, thereby increasing the data volume of the input signal compared to the data volume of the reference signal; an interference sound suppression unit that suppresses the interference sound component included in the input signal using the input signal with the increased data volume and the acquired reference signal; a data deletion unit that restores the data volume of the input signal by deleting the inserted plurality of zeros from the input signal with the suppressed interference sound component; and an output unit that outputs the input signal with the restored data volume.
[0041] (10) Another embodiment of the present invention relates to a non-temporary computer-readable recording medium that records an interference sound suppression program, wherein the interference sound suppression program enables a computer to function as the following units: an input signal acquisition unit that acquires an input signal obtained from a microphone; a reference signal acquisition unit that acquires a reference signal; a data insertion unit that inserts a plurality of zeros into the acquired input signal, thereby increasing the data volume of the input signal compared to the data volume of the reference signal; an interference sound suppression unit that suppresses the interference sound component included in the input signal using the input signal with the increased data volume and the acquired reference signal; a data deletion unit that restores the data volume of the input signal by deleting the inserted plurality of zeros from the input signal with the suppressed interference sound component; and an output unit that outputs the input signal with the restored data volume.
[0042] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. In addition, the following embodiments are only examples of specific implementation of the present disclosure and do not limit the technical scope of the present disclosure.
[0043] (Implementation Method 1)
[0044] Figure 1 1 is a diagram showing a configuration of a communication device 1 in Embodiment 1 of the present disclosure. The communication device 1 is used in a sound-reinforced hands-free communication system, a sound-reinforced two-way communication conference system, an intercom system, and the like mounted in a car or the like.
[0045] Figure 1 The communication device 1 shown includes an echo suppression device 10 , a microphone 11 , an input terminal 12 , a speaker 13 , and an output terminal 14 .
[0046] The microphone 11 is disposed in a space where a talker is present, and collects the voice of the talker. The microphone 11 outputs an input signal representing the collected voice to the echo suppression device 10 .
[0047] The input terminal 12 outputs a reproduced signal received from a communication device (not shown) on the receiving side to the echo suppression device 10 and the speaker 13 .
[0048] The speaker 13 outputs the input reproduction signal to the outside. Here, when the sound output from the speaker 13 is picked up by the microphone 11, the sound of the speaker on the receiving side is reproduced with delay from the speaker on the receiving side, generating a so-called acoustic echo. For this reason, the echo suppression device 10 suppresses the acoustic echo signal included in the input signal output from the microphone 11.
[0049] The output terminal 14 outputs the input signal in which the interfering sound component (acoustic echo component) is suppressed by the echo suppression device 10 .
[0050] The input terminal 12 and the output terminal 14 are connected to a communication unit (not shown). The communication unit transmits an input signal to a communication device (not shown) on the receiving side via a network, and receives a reproduced signal from the communication device (not shown) on the receiving side via the network. The network is, for example, Ethernet.
[0051] Figure 2 1 is a diagram showing a configuration of an echo suppression device 10 in Embodiment 1 of the present disclosure. The echo suppression device 10 is an example of an interfering sound suppression device.
[0052] Figure 2 The echo suppression device 10 shown includes an input signal acquisition unit 101 , a reference signal acquisition unit 102 , a data insertion unit 103 , an echo suppression unit 104 , a data deletion unit 105 , and an output unit 106 .
[0053] The input signal acquisition unit 101 acquires a time domain input signal obtained from the microphone 11. The input signal acquisition unit 101 acquires an input signal with a given sampling frequency and a given number of samples. The given sampling frequency is, for example, 32 kHz, and the given number of samples is, for example, 1024.
[0054] The reference signal acquisition unit 102 acquires a reference signal in the time domain. The reference signal acquisition unit 102 acquires the reproduced signal output to the speaker 13 as a reference signal. The reference signal acquisition unit 102 acquires a reference signal with a given sample frequency and a given number of samples. The given sample frequency is, for example, 32 kHz, and the given number of samples is, for example, 1024.
[0055] The data amount (number of samples) of the input signal acquired by the input signal acquisition unit 101 and the data amount (number of samples) of the reference signal acquired by the reference signal acquisition unit 102 are the same.
[0056] The data inserting unit 103 inserts a plurality of zeros into the input signal acquired by the input signal acquiring unit 101, thereby increasing the data amount of the input signal compared to the data amount of the reference signal. The data inserting unit 103 increases the number of samples of the input signal compared to the number of samples of the reference signal. The data inserting unit 103 increases the number of samples of the input signal to more than twice the number of samples of the reference signal.
[0057] For example, the data inserting unit 103 inserts the same number of zeros as the number of samples of the input signal acquired by the input signal acquiring unit 101 into the input signal. For example, when the number of samples of the input signal acquired by the input signal acquiring unit 101 is 1024, the data inserting unit 103 inserts 1024 zeros into the input signal. As a result, the amount of data of the input signal is doubled, and the number of samples of the input signal is doubled to 2048.
[0058] In addition, the data inserting unit 103 may insert a plurality of zeros after the acquired input signal, or may insert a plurality of zeros before the acquired input signal. In addition, the data inserting unit 103 may also divide the acquired input signal into two, and insert a plurality of zeros between the two divided input signals. In this way, the position where the plurality of zeros are inserted is not limited. In addition, the data inserting unit 103 stores the insertion position of the plurality of zeros in a memory not shown.
[0059] In addition, all of the multiple zeros inserted into the input signal may be values close to 0 such as 0.1 (substantially 0), or some of the multiple zeros inserted into the input signal may be values close to 0 such as 0.1 (substantially 0).
[0060] The echo suppression unit 104 is an example of an interfering sound suppression unit and suppresses an acoustic echo component (interfering sound component) included in the input signal using the input signal with increased data volume by the data inserting unit 103 and the reference signal acquired by the reference signal acquiring unit 102 .
[0061] The echo suppression unit 104 includes a first frequency domain transform unit 111 , a second frequency domain transform unit 112 , a sample number reduction unit 113 , a coefficient calculation unit 114 , a sample number increase unit 115 , a multiplication unit 116 , and a time domain transform unit 117 .
[0062] The first frequency domain conversion unit 111 converts the input signal in the time domain whose data amount has been increased by the data insertion unit 103 into an input signal in the frequency domain by rapidly performing a fast Fourier transform (FFT) of a discrete Fourier transform (DFT).
[0063] The second frequency domain transformer 112 transforms the reference signal in the time domain acquired by the reference signal acquirer 102 into a reference signal in the frequency domain by fast Fourier transform (FFT).
[0064] The sample number reduction unit 113 reduces the number of samples of the input signal in the frequency domain transformed by the first frequency domain transformation unit 111 to the same number of samples as the reference signal. When the number of samples of the input signal is 2048 and the number of samples of the reference signal is 1024, the sample number reduction unit 113 reduces the number of samples of the input signal to half, 1024. For example, the sample number reduction unit 113 reduces the number of samples of the input signal to half by setting the average of the values of two adjacent sample points of the input signal to the value of one sample point. Thus, the data amount (number of samples) of the input signal and the data amount (number of samples) of the reference signal are the same.
[0065] The coefficient calculation unit 114 calculates a suppression coefficient for suppressing an acoustic error component (interference sound component) included in the input signal based on the frequency domain input signal whose sample number is reduced by the sample number reduction unit 113 and the frequency domain reference signal transformed by the second frequency domain transformation unit 112 .
[0066] The coefficient calculation unit 114 calculates the estimated value of the frequency characteristic of the acoustic coupling amount based on the ratio of the power of the input signal to the power of the reference signal. More specifically, the coefficient calculation unit 114 calculates the estimated value of the frequency characteristic of the acoustic coupling amount as the smaller one of the ratio of the power of the input signal to the power of the reference signal and the estimated value of the frequency characteristic of the acoustic coupling amount one frame ago. The coefficient calculation unit 114 calculates the power estimated value of the acoustic echo using the power of the reference signal, the estimated value of the acoustic coupling amount, and the power estimated value of the acoustic echo one frame ago. At this time, the coefficient calculation unit 114 calculates the power estimated value of the acoustic echo by adding the value obtained by multiplying the power of the reference signal by the estimated value of the acoustic coupling amount and the value obtained by multiplying the power estimated value of the acoustic echo one frame ago by the forgetting coefficient matched with the reverberation time.
[0067] The coefficient calculation unit 114 calculates the suppression coefficient using the power of the input signal and the power estimation value of the acoustic echo based on the Wiener filter method. At this time, the coefficient calculation unit 114 calculates the suppression coefficient by dividing the value obtained by subtracting the power estimation value of the acoustic echo from the power of the input signal by the power of the input signal. The suppression coefficient is a real value between 0 and 1. If the input signal contains a large number of acoustic echo components, the value of the suppression coefficient is small, and if the input signal contains a small number of acoustic echo components, the value of the suppression coefficient is large.
[0068] The sample number increasing unit 115 increases the sample number of the suppression coefficient calculated by the coefficient calculating unit 114 to the same sample number as the input signal whose data amount is increased by the data inserting unit 103. When the sample number of the suppression coefficient is 1024 and the sample number of the input signal is 2048, the sample number increasing unit 115 increases the sample number of the suppression coefficient to twice 2048. For example, the sample number increasing unit 115 increases the sample number of the suppression coefficient to twice by setting the value of one sample point of the suppression coefficient to the values of two adjacent sample points. Thus, the data amount (sample number) of the input signal and the data amount (sample number) of the suppression coefficient are the same.
[0069] The multiplication unit 116 multiplies the input signal in the frequency domain transformed by the first frequency domain transformation unit 111 by the suppression coefficient whose number of samples is increased by the sample number increase unit 115. The multiplication unit 116 suppresses the acoustic echo component (interference sound component) included in the input signal by convolving the suppression coefficient with the input signal. Thus, an input signal with suppressed acoustic echo components is generated.
[0070] Also, when performing a convolution operation of convolving the suppression coefficient and the input signal, aliasing distortion occurs. However, in the first embodiment, since a plurality of zeros are inserted into the input signal, the influence of the aliasing distortion generated when performing the convolution operation can be reduced.
[0071] The time domain transform unit 117 transforms the input signal in the frequency domain multiplied by the suppression coefficient by the multiplication unit 116 into an input signal in the time domain by rapidly performing an inverse fast Fourier transform (IFFT) of an inverse discrete Fourier transform (IDFT).
[0072] The data deletion unit 105 deletes the plurality of zeros inserted by the data insertion unit 103 from the input signal whose acoustic echo component (interference sound component) has been suppressed by the echo suppression unit 104, thereby restoring the data amount of the input signal. In addition, a memory (not shown) stores the insertion positions of the plurality of zeros inserted by the data insertion unit 103. The data deletion unit 105 deletes the plurality of zeros inserted by the data insertion unit 103 from the input signal in the time domain with reference to the insertion positions of the plurality of zeros stored in the memory (not shown).
[0073] The output unit 106 outputs the input signal whose data volume has been restored by the data deleting unit 105 to the output terminal 14 .
[0074] Next, the operation of the echo suppression device 10 in the first embodiment of the present disclosure will be described.
[0075] Figure 3 This is a flowchart for explaining the operation of the echo suppression device 10 in the first embodiment of the present disclosure.
[0076] First, in step S1 , the input signal acquisition unit 101 acquires an input signal in the time domain from the microphone 11 .
[0077] Next, in step S2, the reference signal acquisition unit 102 acquires a reference signal in the time domain. The acquired reference signal is a reproduction signal output to the speaker 13. In addition, the number of samples of the input signal acquired by the input signal acquisition unit 101 is the same as the number of samples of the reference signal acquired by the reference signal acquisition unit 102.
[0078] Next, in step S3, the data inserting unit 103 inserts a plurality of zeros into the input signal acquired by the input signal acquiring unit 101. At this time, the data inserting unit 103 inserts the same number of zeros into the input signal as the number of samples of the input signal acquired by the input signal acquiring unit 101. Thus, the data amount of the input signal becomes twice the data amount of the reference signal.
[0079] Next, in step S4, the first frequency domain conversion unit 111 converts the input signal in the time domain, the amount of which is increased by the data insertion unit 103, into an input signal in the frequency domain. At this time, the first frequency domain conversion unit 111 performs a fast Fourier transform on the input signal in the time domain, the amount of which is increased by the data insertion unit 103. Thus, the input signal in the time domain is converted into an input signal in the frequency domain. The first frequency domain conversion unit 111 outputs the converted input signal in the frequency domain to the sample number reduction unit 113 and the multiplication unit 116.
[0080] Next, in step S5, the second frequency domain transform unit 112 transforms the reference signal in the time domain acquired by the reference signal acquisition unit 102 into a reference signal in the frequency domain. At this time, the second frequency domain transform unit 112 performs a fast Fourier transform on the reference signal in the time domain acquired by the reference signal acquisition unit 102. Thus, the reference signal in the time domain is transformed into a reference signal in the frequency domain.
[0081] Next, in step S6, the sample number reduction unit 113 reduces the number of samples of the input signal in the frequency domain transformed by the first frequency domain transformation unit 111 to the same number of samples as the reference signal. For example, the sample number reduction unit 113 sets the number of samples of the input signal in the frequency domain to 1 / 2. For example, the sample number reduction unit 113 sets the average of the values of two adjacent sample points of the input signal to the value of one sample point. Thus, the data amount (number of samples) of the input signal and the data amount (number of samples) of the reference signal are the same.
[0082] Next, in step S7, the coefficient calculation unit 114 calculates a suppression coefficient for suppressing an acoustic error component (interference sound component) contained in the input signal based on the input signal in the frequency domain and the reference signal in the frequency domain. The suppression coefficient is, for example, a Wiener filter. For example, the coefficient calculation unit 114 calculates the suppression coefficient based on the input signal in the frequency domain and the reference signal in the frequency domain using the Wiener filter method.
[0083] Next, in step S8, the sample number increasing unit 115 increases the sample number of the suppression coefficient calculated by the coefficient calculating unit 114 by the same number of samples as the input signal whose data amount (sample number) has been increased. For example, the sample number increasing unit 115 sets the sample number of the suppression coefficient calculated by the coefficient calculating unit 114 to two times. For example, the sample number increasing unit 115 sets the value of one sample point of the suppression coefficient to the values of two adjacent sample points. Thus, the data amount (sample number) of the suppression coefficient is the same as the data amount (sample number) of the input signal.
[0084] Next, in step S9 , the multiplication unit 116 multiplies the input signal in the frequency domain transformed by the first frequency domain transformation unit 111 by the suppression coefficient whose number of samples is increased by the sample number increase unit 115 .
[0085] Next, in step S10, the time domain conversion unit 117 converts the input signal in the frequency domain multiplied by the suppression coefficient by the multiplication unit 116 into an input signal in the time domain. At this time, the time domain conversion unit 117 performs an inverse fast Fourier transform on the input signal in the frequency domain multiplied by the suppression coefficient by the multiplication unit 116. Thus, the input signal in the frequency domain is converted into an input signal in the time domain.
[0086] Next, in step S11, the data deleting unit 105 deletes the plurality of zeros inserted by the data inserting unit 103 from the input signal in the time domain in which the acoustic echo component (interference sound component) has been suppressed after transformation by the time domain transforming unit 117. Thus, the data amount of the input signal is restored.
[0087] Next, in step S12 , the output unit 106 outputs the input signal whose data volume has been restored by the data deleting unit 105 to the output terminal 14 .
[0088] In this way, by inserting multiple zeros into the input signal, the influence of aliasing distortion generated when suppressing the acoustic error component (interference sound component) included in the input signal can be reduced, thereby improving the call quality. In addition, by inserting multiple zeros only into the input signal and not inserting multiple zeros into the reference signal, the amount of calculation can be reduced compared to the case where multiple zeros are inserted into both the input signal and the reference signal.
[0089] Furthermore, in the first embodiment, the communication device 1 includes one speaker 13 and one output terminal 14 , but the present disclosure is not particularly limited thereto, and the communication device 1 may include a plurality of speakers 13 and a plurality of output terminals 14 .
[0090] In this case, the reference signal acquisition unit 102 may acquire a plurality of reference signals. The reference signal acquisition unit 102 may acquire a plurality of reproduced signals output to the plurality of speakers 13 as a plurality of time domain reference signals. The echo suppression unit 104 may suppress the acoustic echo component (interference sound component) contained in the input signal using the input signal whose data volume has been increased by the data insertion unit 103 and the plurality of reference signals acquired by the reference signal acquisition unit 102. The second frequency domain transformation unit 112 may transform the plurality of time domain reference signals acquired by the reference signal acquisition unit 102 into a plurality of frequency domain reference signals by fast Fourier transform (FFT). The coefficient calculation unit 114 may calculate power estimation values of a plurality of acoustic echoes corresponding to the plurality of reference signals, and calculate the suppression coefficient using the power of the input signal and the power estimation values of the plurality of acoustic echoes.
[0091] In the first embodiment, the communication device 1 includes one microphone 11 and one input terminal 12, but the present disclosure is not particularly limited thereto, and the communication device 1 may include multiple microphones 11 and multiple input terminals 12. In this case, the communication device 1 includes multiple echo suppression devices 10 according to the number of the multiple microphones 11 and the multiple input terminals 12.
[0092] (Implementation Method 2)
[0093] The echo suppression device 10 in the first embodiment suppresses the acoustic echo signal included in the input signal output from the microphone 11. In contrast, the noise suppression device in the second embodiment suppresses the noise included in the input signal output from the microphone 11.
[0094] Figure 4 2 is a diagram showing a configuration of a voice processing device 2 in Embodiment 2 of the present disclosure. The voice processing device 2 is used in a voice recognition system, a speaker recognition system, and the like.
[0095] Figure 4 The illustrated sound processing device 2 includes a noise suppression device 10A, a first microphone 11A, a second microphone 11B, and an output terminal 14 .
[0096] The first microphone 11A is arranged in a space where a speaker is present, and collects the speaker's voice. The first microphone 11A outputs an input signal representing the collected voice to the noise suppression device 10A. The first microphone 11A has the same configuration as the microphone 11 in the first embodiment.
[0097] The second microphone 11B is arranged in a space where the talker is present, and collects noise generated from a noise source existing around the talker. The second microphone 11B outputs a noise signal representing the collected noise as a reference signal to the noise suppression device 10A.
[0098] If there is a noise source around the speaker, the sound picked up by the first microphone 11A may contain noise. Therefore, the noise suppression device 10A suppresses the noise signal contained in the input signal output from the first microphone 11A.
[0099] The output terminal 14 outputs the input signal from which the interfering sound component (noise component) has been suppressed by the noise suppression device 10A.
[0100] In addition, the output terminal 14 is connected to a voice recognition device or a speaker recognition device. The voice recognition device recognizes the voice of the speaker using the input signal with the noise component suppressed output from the output terminal 14 of the voice processing device 2. In addition, the speaker recognition device recognizes the speaker using the input signal with the noise component suppressed output from the output terminal 14 of the voice processing device 2.
[0101] Figure 5 1 is a diagram showing a configuration of a noise suppression device 10A in Embodiment 2 of the present disclosure. The noise suppression device 10A is an example of an interfering sound suppression device.
[0102] Figure 5 The noise suppression device 10A shown in the figure includes an input signal acquisition unit 101, a reference signal acquisition unit 102, a data insertion unit 103, a noise suppression unit 104A, a data deletion unit 105, and an output unit 106. In addition, in the second embodiment, the same reference numerals are given to the same structures as those in the first embodiment, and the description thereof is omitted. In the following description, the points different from the first embodiment are described.
[0103] The input signal acquisition unit 101 acquires a time domain input signal obtained from the first microphone 11A. The input signal acquisition unit 101 acquires an input signal of a given sampling frequency and a given number of samples. The given sampling frequency is, for example, 32 kHz, and the given number of samples is, for example, 1024.
[0104] The reference signal acquisition unit 102 acquires a reference signal in the time domain. The reference signal is a noise signal obtained from a second microphone 11B different from the first microphone 11A. The reference signal acquisition unit 102 acquires the noise signal output from the second microphone 11B as a reference signal. The reference signal acquisition unit 102 acquires a reference signal of a given sample frequency and a given number of samples. The given sample frequency is, for example, 32 kHz, and the given number of samples is, for example, 1024.
[0105] The noise suppression unit 104A is an example of an interference sound suppression unit and uses the input signal with increased data volume by the data inserting unit 103 and the reference signal acquired by the reference signal acquiring unit 102 to suppress the noise component (interference sound component) included in the input signal.
[0106] The noise suppression unit 104A includes a first frequency domain transform unit 111 , a second frequency domain transform unit 112 , a sample number reduction unit 113 , a coefficient calculation unit 114 , a sample number increase unit 115 , a multiplication unit 116 , and a time domain transform unit 117 .
[0107] The coefficient calculation unit 114 calculates a suppression coefficient for suppressing a noise component (interference sound component) included in the input signal based on the input signal in the frequency domain whose number of samples is reduced by the sample number reduction unit 113 and the reference signal in the frequency domain transformed by the second frequency domain transformation unit 112. The method for calculating the suppression coefficient by the coefficient calculation unit 114 in the second embodiment is the same as the method for calculating the suppression coefficient by the coefficient calculation unit 114 in the first embodiment. The coefficient calculation unit 114 in the second embodiment calculates a power estimation value of noise, and calculates the suppression coefficient using the power of the input signal and the power estimation value of noise based on the Wiener filter method.
[0108] The data deleting unit 105 deletes the plurality of zeros inserted by the data inserting unit 103 from the input signal whose noise component (interference sound component) has been suppressed by the noise suppressing unit 104A, thereby restoring the data amount of the input signal.
[0109] Note that the operation of the noise suppression device 10A in the second embodiment is the same as the operation of the echo suppression device 10 in the first embodiment, and therefore the description thereof will be omitted.
[0110] In this way, since the reference signal is a noise signal obtained from the second microphone 11B different from the first microphone 11A, it is possible to suppress the noise component included in the input signal.
[0111] In addition, when the direction from which the speaker's voice comes and the direction from which the noise comes are known, the individual sounds can be separated by beamforming technology. To this end, the sound processing device 2 may also include: a noise suppression device 10A, a microphone array, a second microphone 11B, an output terminal 14, and a beamformer. The microphone array is composed of a plurality of microphones. The microphone array outputs the sound signal picked up to the beamformer. The beamformer may also separate the sound signal from the microphone array into a first sound signal coming from the first direction and a second sound signal coming from the second direction when the first direction from which the speaker's voice comes and the second direction from which the noise comes are known. In addition, the beamformer may also output the first sound signal as an input signal to the noise suppression device 10A, and output the second sound signal as a reference signal to the noise suppression device 10A. The input signal acquisition unit 101 may also acquire the input signal obtained from the microphone array. The reference signal acquisition unit 102 may also acquire the noise signal from a given direction obtained from the microphone array as a reference signal.
[0112] In addition, in each of the above-mentioned embodiments, each structural element may also be constituted by dedicated hardware, or implemented by executing a software program suitable for each structural element. Each structural element may also be implemented by a program execution unit such as a CPU or a processor reading and executing a software program recorded in a recording medium such as a hard disk or a semiconductor memory. In addition, the program may also be implemented by another independent computer system by recording the program in a recording medium and transferring it, or transferring the program via a network.
[0113] Part or all of the functions of the device involved in the embodiments of the present disclosure are typically implemented as LSI (Large Scale Integration) as an integrated circuit. They can be single-chip (onechip) alone, or they can include a part or all of the single-chip. In addition, integrated circuitization is not limited to LSI, and can also be implemented by a dedicated circuit or a general-purpose processor. It is also possible to use an FPGA (Field Programmable Gate Array) that can be programmed after LSI manufacturing, or a reconfigurable processor that can reconfigure the connection and settings of the circuit monomers inside the LSI.
[0114] Furthermore, a part or all of the functions of the apparatus according to the embodiments of the present disclosure may be realized by executing a program by a processor such as a CPU.
[0115] In addition, the numbers used above are all exemplified numbers in order to specifically describe the present disclosure, and the present disclosure is not limited to the exemplified numbers.
[0116] In addition, the order in which the steps shown in the above flowchart are executed is an example order for specifically explaining the present disclosure, and can be an order other than the above order within the scope of being able to obtain the same effect. In addition, part of the above steps can also be executed simultaneously (in parallel) with other steps.
[0117] Industrial Applicability
[0118] The technology according to the present disclosure can improve call quality and reduce the amount of calculation, and is therefore useful as a technology for suppressing interference sounds included in an input signal obtained from a microphone.
Claims
1. An interference sound suppression device, comprising: An input signal acquisition unit, which acquires an input signal obtained from a microphone; A reference signal acquisition unit, which acquires a reference signal; a data inserting unit, which inserts a plurality of zeros into the acquired input signal, thereby increasing the data amount of the input signal compared to the data amount of the reference signal; an interference sound suppression unit, using the input signal with an increased data volume and the acquired reference signal to suppress an interference sound component included in the input signal; a data deleting unit that restores the data amount of the input signal by deleting the inserted plurality of zeros from the input signal in which the interfering sound component is suppressed; and The output unit outputs the input signal with the data amount restored.
2. The interference sound suppression device according to claim 1, wherein: The data inserting unit increases the number of samples of the input signal to be twice or more the number of samples of the reference signal.
3. The interference sound suppression device according to claim 1 or 2, wherein: The reference signal is a reproduced signal output to a speaker.
4. The interference sound suppression device according to claim 1 or 2, wherein: The reference signal is a noise signal obtained from a microphone other than the microphone.
5. The interference sound suppression device according to claim 1 or 2, wherein: The reference signal is a noise signal obtained from the microphone from a given direction.
6. The interference sound suppression device according to claim 1 or 2, wherein: The input signal acquisition unit acquires the input signal in the time domain. The reference signal acquisition unit acquires the reference signal in the time domain, The interference sound suppression unit comprises: a first frequency domain conversion unit that converts the input signal in the time domain with an increased data amount into an input signal in the frequency domain; A second frequency domain transform unit, transforming the acquired reference signal in the time domain into a reference signal in the frequency domain; a sample number reducing unit, which reduces the number of samples of the transformed input signal in the frequency domain to the same number of samples as that of the reference signal; a coefficient calculation unit that calculates a suppression coefficient for suppressing the interfering sound component based on the input signal in the frequency domain whose number of samples is reduced and the reference signal in the frequency domain after the transformation; a sample number increasing unit that increases the sample number of the calculated suppression coefficient to the same sample number as the input signal with the increased data amount; a multiplication unit for multiplying the transformed input signal in the frequency domain by the suppression coefficient for increasing the number of samples; and The time domain transform unit transforms the input signal in the frequency domain multiplied by the suppression coefficient into an input signal in the time domain.
7. The interference sound suppression device according to claim 1 or 2, wherein: The reference signal acquisition unit acquires a plurality of reference signals. The interference sound suppression unit suppresses the interference sound component included in the input signal using the input signal with an increased data amount and the acquired plurality of reference signals.
8. A method for suppressing interference sound, which is a method for suppressing interference sound in a computer, comprising: Get the input signal from the microphone, Get the reference signal, By inserting a plurality of zeros into the acquired input signal, the data amount of the input signal is increased compared with the data amount of the reference signal, using the input signal with increased data volume and the acquired reference signal to suppress an interference sound component included in the input signal, The data amount of the input signal is restored by deleting the inserted plurality of zeros from the input signal in which the interfering sound component is suppressed, The input signal with the data amount restored is output.
9. An interference sound suppression program, causing a computer to function as the following units: An input signal acquisition unit, which acquires an input signal obtained from a microphone; A reference signal acquisition unit, which acquires a reference signal; a data inserting unit, which inserts a plurality of zeros into the acquired input signal, thereby increasing the data amount of the input signal compared to the data amount of the reference signal; an interference sound suppression unit, using the input signal with an increased data volume and the acquired reference signal to suppress an interference sound component included in the input signal; a data deleting unit that restores the data amount of the input signal by deleting the inserted plurality of zeros from the input signal in which the interfering sound component is suppressed; and The output unit outputs the input signal with the data amount restored.
