A microphone array sound source separation method and system with gain linkage parameter update

Abstract

The present invention relates to a microphone array sound source separation method and system with gain linkage parameter update, which belongs to the field of sound source separation. The method comprises: obtaining an observation signal of a sound source by using a microphone array; constructing an observation signal vector according to the observation signal; constructing a beamformer; constructing a matrix according to the beamformer; obtaining a gain parameter; calculating a gain function according to the matrix and the gain parameter; updating the matrix according to the gain function to obtain an update matrix; determining a sound source signal after multi-path separation according to the update matrix and the observation signal vector, that is, realizing sound source separation. The present invention can realize the separation of sound signals.

Description

Technical Field

[0001] The present invention relates to the technical field of sound source separation, and in particular to a microphone array sound source separation method and system with gain linkage parameter update. Background Art

[0002] When using a microphone array to observe sound source signals, various types of noise and interference are often observed in the surrounding area. Since there are many types of interference, it is necessary to use the proposed sound source separation method to separate the sound signals. Summary of the invention

[0003] The object of the present invention is to provide a microphone array sound source separation method and system with gain linkage parameter update, which can realize the separation of sound signals.

[0004] To achieve the above object, the present invention provides the following solutions:

[0005] A microphone array sound source separation method with gain linkage parameter update, comprising:

[0006] Using a microphone array to obtain observation signals of sound sources;

[0007] Constructing an observation signal vector according to the observation signal;

[0008] Build a beamformer;

[0009] constructing a matrix according to the beamformer;

[0010] Get gain parameters;

[0011] Calculate a gain function according to the matrix and the gain parameter;

[0012] Update the matrix according to the gain function to obtain an updated matrix;

[0013] The sound source signal after multi-path separation is determined according to the update matrix and the observation signal vector, thus achieving sound source separation.

[0014] Optionally, the observed signal vector is:

[0015] y(f,t)=[Y 0 (f,t) Y 1 (f,t) … Y M-1 (f,t)] T ,

[0016] Where y(f,t) is the observed signal vector, f is the frequency, t is the time, the superscript T represents the transposition, and Y M-1 (f, t) represents the observation signal of the mth microphone in the time-frequency domain, m = 0, 1, 2, ..., M-1, and M is the number of sensors in the microphone array.

[0017] Optionally, the gain function is calculated using the following formula:

[0018]

[0019] Among them, g n,m (f,t) is the gain function, φ n (f, t) is the nth element in the matrix, λ n,m (f) is the energy distribution of the nth sound source in the mth feature space.

[0020] Optionally, the matrix is ​​updated using the following formula:

[0021]

[0022]

[0023]

[0024]

[0025]

[0026] w m (f)←[W(f)B m (f)] -1 i m

[0027]

[0028] in, is the nth frequency parameter in the matrix, is the nth time parameter in the matrix, g n,m (f,t) is the gain function, ξ m (f,t) is the time-frequency parameter, φ n (f, t) is the nth element in the matrix, λ n,m (f) is the energy distribution of the nth sound source in the mth feature space, B m (f) is the square matrix constructed for the observed signal, G m (f, t) is a gate function, y(f, t) is the observed signal vector, y H (f, t) is the conjugate transpose of the observed signal vector, W(f) is the matrix, w m (f) is the mth column of the matrix, i m is the mth column of the identity matrix, w H (f) is the update matrix.

[0029] Optionally, the following formula is used to determine the sound source signal after multi-path separation:

[0030] Z(f,t)=w H (f)y(f,t)

[0031] Among them, Z(f,t) is the separated sound source signal, w H (f) is the update matrix, and y(f,t) is the observed signal vector.

[0032] A microphone array sound source separation system with gain linkage parameter update, comprising:

[0033] An observation signal acquisition module is used to acquire the observation signal of the sound source using a microphone array;

[0034] An observation signal vector construction module, used to construct an observation signal vector according to the observation signal;

[0035] A beamformer building module, used to build a beamformer;

[0036] A matrix construction module, used for constructing a matrix according to the beamformer;

[0037] A gain parameter acquisition module, used to obtain a gain parameter;

[0038] A gain function calculation module, used for calculating the gain function according to the matrix and the gain parameter;

[0039] A matrix updating module, used for updating the matrix according to the gain function to obtain an updated matrix;

[0040] The sound source separation module is used to determine the sound source signal after multi-path separation according to the update matrix and the observation signal vector, that is, to achieve sound source separation.

[0041] Optionally, the observation signal vector constructed by the observation signal vector construction module is:

[0042] y(f,t)=[Y 0 (f,t) Y 1 (f,t) … Y M-1 (f,t)] T ,

[0043] Where y(f,t) is the observed signal vector, f is the frequency, t is the time, the superscript T represents the transposition, and Y M-1 (f, t) represents the observation signal of the mth microphone in the time-frequency domain, m = 0, 1, 2, ..., M-1, and M is the number of sensors in the microphone array.

[0044] Optionally, the gain function calculation module calculates the gain function using the following formula:

[0045]

[0046] Among them, g n,m (f,t) is the gain function, φ n (f, t) is the nth element in the matrix, λ n,m (f) is the energy distribution of the nth sound source in the mth feature space.

[0047] Optionally, the matrix updating module updates the matrix using the following formula:

[0048]

[0049]

[0050]

[0051]

[0052]

[0053] w m (f)←[W(f)B m (f)] -1 i m

[0054]

[0055] in, is the nth frequency parameter in the matrix, is the nth time parameter in the matrix, g n,m (f,t) is the gain function, ξ m (f,t) is the time-frequency parameter, φ n (f, t) is the nth element in the matrix, λ n,m (f) is the energy distribution of the nth sound source in the mth feature space, B m (f) is the square matrix constructed for the observed signal, G m (f, t) is a gate function, y(f, t) is the observed signal vector, y H (f, t) is the conjugate transpose of the observed signal vector, W(f) is the matrix, w m (f) is the mth column of the matrix, i m is the mth column of the identity matrix, w H (f) is the update matrix.

[0056] Optionally, the sound source separation module uses the following formula to determine the sound source signal after multi-path separation:

[0057] Z(f,t)=w H (f)y(f,t)

[0058] Among them, Z(f,t) is the separated sound source signal, w H (f) is the update matrix, and y(f,t) is the observed signal vector.

[0059] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0060] The present invention introduces a gain parameter as a matrix optimization criterion and applies the update matrix to the observed signal vector, thereby obtaining multiple array outputs, each of which corresponds to a separated signal, thereby achieving sound source separation. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0062] Figure 1 A flow chart of a microphone array sound source separation method with gain linkage parameter update according to the present invention;

[0063] Figure 2 This is a module diagram of the microphone array sound source separation system with gain linkage parameter update of the present invention. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] The object of the present invention is to provide a microphone array sound source separation method and system with gain linkage parameter update, which can realize the separation of sound signals.

[0066] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0067] Figure 1 The flowchart of the microphone array sound source separation method with gain linkage parameter update of the present invention is as follows: Figure 1 As shown, a microphone array sound source separation method with gain linkage parameter update includes:

[0068] Step 101: Acquire an observation signal of a sound source using a microphone array.

[0069] Step 102: construct an observation signal vector according to the observation signal.

[0070] Specifically:

[0071] Use a microphone array to observe the sound source signal, and list all observed signals into an observation signal vector.

[0072] Assume that the observation signal of the mth (m=0,1,2,…,M-1) microphone in the time-frequency domain is Y m (f,t), where f is frequency and t is time.

[0073] List all observations into a vector, the observation signal vector:

[0074] y(f,t)=[Y 0 (f,t) Y 1 (f,t) … Y M-1 (f,t)] T ,

[0075] The superscript T stands for transpose.

[0076] Step 103: Construct a beamformer.

[0077] Specifically, a similar method as described above is used to construct another vector to be optimized, namely the beamformer w(f, t). The length of this vector is also M, which is the same as the length of the observed signal vector.

[0078] Step 104: construct a matrix according to the beamformer.

[0079] If there are I beamformers w i (f), we can construct an M×I matrix, expressed as:

[0080] W(f)=[w 0 (f) w 1 (f) … w I-1 (f)],

[0081] Step 105: Obtain gain parameters.

[0082] Specifically, the gain parameters include φ n (f,t) and λ n,m (f).

[0083] Step 106: Calculate a gain function according to the matrix and the gain parameter.

[0084] Specifically, calculate the gain function g n,m(f, t) uses the following formula:

[0085]

[0086] Among them, g n,m (f,t) is the gain function, φ n (f, t) is the nth element in the matrix, λ n,m (f) is the energy distribution of the nth sound source in the mth feature space.

[0087] Step 107: Update the matrix according to the gain function to obtain an updated matrix.

[0088] The update process is:

[0089]

[0090]

[0091]

[0092]

[0093]

[0094] w m (f)←[W(f)B m (f)] -1 i m

[0095]

[0096] in, is the nth frequency parameter in the matrix, is the nth time parameter in the matrix, g n,m (f,t) is the gain function, ξ m (f,t) is the time-frequency parameter, φ n (f, t) is the nth element in the matrix, λ n,m (f) is the energy distribution of the nth sound source in the mth feature space, B m (f) is the square matrix constructed for the observed signal, G m (f, t) is a gate function, y(f, t) is the observed signal vector, y H (f, t) is the conjugate transpose of the observed signal vector, W(f) is the matrix, w m (f) is the mth column of the matrix, i m is the mth column of the identity matrix, w H (f) is the update matrix.

[0097] Step 108: Determine the sound source signal after multi-path separation according to the update matrix and the observed signal vector, that is, realize sound source separation.

[0098] After the algorithm converges in step 107, the update matrix is ​​applied to the observed signal vector to obtain multiple array outputs, each of which corresponds to a separated signal, thus achieving sound source separation.

[0099] Specifically:

[0100] Z(f,t)=w H (f)y(f,t)

[0101] z(f,t)=[Z 0 (f,t) Z 1 (f,t) … Z I-1 (f,t)].

[0102] Among them, Z(f,t) is the separated sound source signal, and z(f,t) is the signal vector composed of the separated sound source signals.

[0103] Based on the above method, the present invention also discloses a microphone array sound source separation system with gain linkage parameter update, such as Figure 2 As shown, including:

[0104] The observation signal acquisition module 201 is used to acquire the observation signal of the sound source using a microphone array.

[0105] The observation signal vector construction module 202 is used to construct an observation signal vector according to the observation signal.

[0106] The beamformer construction module 203 is used to construct a beamformer.

[0107] The matrix construction module 204 is configured to construct a matrix according to the beamformer.

[0108] The gain parameter acquisition module 205 is used to acquire the gain parameter.

[0109] The gain function calculation module 206 is used to calculate the gain function according to the matrix and the gain parameter.

[0110] The matrix updating module 207 is used to update the matrix according to the gain function to obtain an updated matrix.

[0111] The sound source separation module 208 is used to determine the multi-path separated sound source signals according to the update matrix and the observed signal vector, that is, to achieve sound source separation.

[0112] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0113] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A microphone array sound source separation method with gain linkage parameter update, It is characterized in that include: Using a microphone array to obtain observation signals of sound sources; An observation signal vector is constructed according to the observation signal; the observation signal vector is: y(f,t)=[Y 0 (f,t) Y 1 (f,t) … Y M-1 (f,t)] T , Where y(f,t) is the observed signal vector, f is the frequency, t is the time, the superscript T represents the transposition, and Y M-1 (f, t) represents the observation signal of the mth microphone in the time-frequency domain, m = 0, 1, 2, ..., M-1, M is the number of sensors in the microphone array; Build a beamformer; constructing a matrix according to the beamformer; Get gain parameters; Calculate a gain function according to the matrix and the gain parameter; The matrix is ​​updated according to the gain function to obtain an updated matrix; the matrix is ​​updated using the following formula: in, is the nth frequency parameter in the matrix, is the nth time parameter in the matrix, g n,m (f,t) is the gain function, ξ m (f,t) is the time-frequency parameter, φ n (f, t) is the nth element in the matrix, λ n,m (f) is the energy distribution of the nth sound source in the mth feature space, B m (f) is the square matrix constructed for the observed signal, G m (f, t) is a gate function, y(f, t) is the observed signal vector, y H (f, t) is the conjugate transpose of the observed signal vector, W(f) is the matrix, w m (f) is the mth column of the matrix, i m is the mth column of the identity matrix, w H (f) is the update matrix; The sound source signal after multi-path separation is determined according to the update matrix and the observation signal vector, thus achieving sound source separation.

2. The microphone array sound source separation method with gain linkage parameter update according to claim 1, It is characterized in that The gain function is calculated using the following formula: Among them, g n,m (f,t) is the gain function, φ n (f, t) is the nth element in the matrix, λ n,m (f) is the energy distribution of the nth sound source in the mth feature space.

3. The microphone array sound source separation method with gain linkage parameter update according to claim 1, It is characterized in that The following formula is used to determine the sound source signal after multi-path separation: Z(f,t)=w H (f)y(f,t) Among them, Z(f,t) is the separated sound source signal, w H (f) is the update matrix, and y(f,t) is the observed signal vector.

4. A microphone array sound source separation system with gain linkage parameter update, It is characterized in that include: An observation signal acquisition module is used to acquire the observation signal of the sound source using a microphone array; An observation signal vector construction module is used to construct an observation signal vector according to the observation signal; the observation signal vector constructed by the observation signal vector construction module is: y(f,t) = [Y 0 (f,t) Y 1 (f,t) … Y M-1 (f,t)] T , Where y(f,t) is the observed signal vector, f is the frequency, t is the time, the superscript T represents the transposition, and Y M-1 (f, t) represents the observation signal of the mth microphone in the time-frequency domain, m = 0, 1, 2, ..., M-1, M is the number of sensors in the microphone array; A beamformer building module, used to build a beamformer; A matrix construction module, used for constructing a matrix according to the beamformer; A gain parameter acquisition module, used to obtain gain parameters; A gain function calculation module, used for calculating the gain function according to the matrix and the gain parameter; A matrix updating module is used to update the matrix according to the gain function to obtain an updated matrix; the matrix updating module uses the following formula to update the matrix: in, is the nth frequency parameter in the matrix, is the nth time parameter in the matrix, g n,m (f,t) is the gain function, ξ m (f,t) is the time-frequency parameter, φ n (f, t) is the nth element in the matrix, λ n,m (f) is the energy distribution of the nth sound source in the mth feature space, B m (f) is the square matrix constructed for the observed signal, G m (f, t) is a gate function, y(f, t) is the observed signal vector, y H (f, t) is the conjugate transpose of the observed signal vector, W(f) is the matrix, w m (f) is the mth column of the matrix, i m is the mth column of the identity matrix, w H (f) is the update matrix; The sound source separation module is used to determine the multi-path separated sound source signal according to the update matrix and the observation signal vector, that is, to achieve sound source separation.

5. The microphone array sound source separation system with gain linkage parameter update according to claim 4, It is characterized in that The gain function calculation module uses the following formula to calculate the gain function: Among them, g n,m (f,t) is the gain function, φ n (f, t) is the nth element in the matrix, λ n,m (f) is the energy distribution of the nth sound source in the mth feature space.

6. The microphone array sound source separation system with gain linkage parameter update according to claim 4, It is characterized in that The sound source separation module uses the following formula to determine the sound source signal after multi-path separation: Z(f,t)=w H (f)y(f,t) Among them, Z(f,t) is the separated sound source signal, w H (f) is the update matrix, and y(f,t) is the observed signal vector.

Images