Sound separation method and system based on vector microphone

Abstract

The invention relates to the field of audio processing, in particular to a sound separation method and system based on vector microphones, and the method comprises the steps: constructing a fourth-order cumulant matrix of sound signals received by the vector microphones, and solving sound source direction angle estimation and sound source pitch angle estimation of each incident signal; constructing a spatial characteristic matrix of each incident signal at different frequencies; based on the spatial characteristic matrix, calculating spatial correlation matrix estimation of the sound signal under different frequencies and different frame numbers, and constructing a spatial correlation matrix of the sound signal under different frequencies and different frame numbers; an orthogonalized MNMF model is constructed; adjusting the variable of the MNMF through a multiplication updating method, enabling the difference between the spatial correlation matrix estimation and the spatial correlation matrix to be smaller than a preset target, and obtaining the optimal spatial correlation matrix estimation; and separating the sound signals based on the optimal spatial correlation matrix. According to the invention, the problem of sound separation when the sound source spacing is short is effectively solved, and the accuracy of sound separation is improved.

Description

Technical Field

[0001] The present invention relates to the field of audio processing, and in particular to a sound separation method and system based on a vector microphone. Background Art

[0002] Sound separation is a signal processing technology that aims to separate a single sound source from a mixed sound signal. This technology plays an important role in daily life and work. It can not only be used to improve the performance of hearing aids and noise-canceling headphones, but also help improve the experience and effects of teleconferencing, speech recognition, music production and many other fields. Simply put, sound separation technology allows people to clearly hear the sounds they want to focus on in a noisy environment, which is crucial for people's communication and understanding.

[0003] In the prior art, technicians usually perform sound separation based on the independent component analysis (ICA) method, the basic principle of which is based on the statistical independence of signals, and blind source separation is achieved by maximizing the statistical independence of output signals. However, when the environment becomes complicated, such as when there is Gaussian noise or the source signal spacing is close, the performance of ICA will be affected. Gaussian noise will interfere with the reception of signals and reduce the signal quality, while the close spacing of source signals may cause signal overlap and confusion, which will cause problems in the accuracy of ICA sound separation. Summary of the invention

[0004] In order to solve the above problems, the present invention provides a sound separation method and system based on vector microphone.

[0005] The first aspect of the present invention discloses a sound separation method based on a vector microphone, comprising:

[0006] receiving, using a vector microphone, an acoustic signal comprising a plurality of incident signals;

[0007] Construct a fourth-order cumulant matrix of the sound signal, and solve the sound source direction angle estimation and sound source pitch angle estimation of each incident signal based on the fourth-order cumulant matrix;

[0008] Based on the estimation of the sound source direction angle and the sound source elevation angle, the spatial characteristic matrix of each incident signal at different frequencies is constructed;

[0009] Based on the spatial characteristic matrix, the spatial correlation matrix estimation of the sound signal at different frequencies and different frame numbers is calculated, and the spatial correlation matrix of the sound signal at different frequencies and different frame numbers is constructed;

[0010] Constructing an orthogonalized MNMF model based on the spatial characteristic matrix, the spatial correlation matrix estimate and the spatial correlation matrix;

[0011] The variables of MNMF are adjusted by the multiplication update method, and the spatial correlation matrix is ​​gradually updated so that the difference between the spatial correlation matrix estimate and the spatial correlation matrix is ​​less than the preset target, and the optimal spatial correlation matrix estimate is obtained;

[0012] Separate the sound signals based on the optimal spatial correlation matrix.

[0013] Furthermore, the sound signal is the sum of the target signal and the noise, the target signal is the sum of the products of each incident signal and its guide vector; the noise includes sound pressure noise and vibration noise on the x, y and z axes.

[0014] Furthermore, the steps of solving the sound source direction angle estimation and the sound source elevation angle estimation of each incident signal based on the fourth-order cumulant matrix include:

[0015] Perform eigendecomposition on the fourth-order cumulant matrix to obtain the signal subspace and noise subspace of the sound signal;

[0016] Calculate multiple sound source direction angles and multiple sound source elevation angles of each incident signal based on the noise subspace;

[0017] According to the correlation between the signal subspace and the noise subspace, the sound source direction angle estimation and the sound source elevation angle estimation of the corresponding incident signal are screened out from multiple sound source direction angles and elevation angles.

[0018] Furthermore, the spatial characteristic matrix of each incident signal at different frequencies is constructed. For the spatial characteristic matrix H of the pth incident signal at frequency i ip for:

[0019]

[0020] in, is the sound source direction angle estimate of the pth incident signal, is the estimated sound source elevation angle of the pth incident signal.

[0021] Furthermore, based on the spatial characteristic matrix, the step of estimating the spatial correlation matrix of the sound signal at different frequencies and different frame numbers includes:

[0022] Spatial correlation matrix estimation at the i-th frequency and frame number l

[0023]

[0024] Among them, t ik is the amplitude of the kth incident signal at frequency i, v kl is the gain of the short-time Fourier transform of the kth incident signal under the frame number l, K represents the number of incident signals, H ikRepresents the spatial characteristic matrix of the kth incident signal at frequency i.

[0025] Furthermore, the steps of constructing the spatial correlation matrix of the sound signal at different frequencies and different frame numbers include:

[0026] The spatial correlation matrix X at the i-th frequency and frame number l il :

[0027]

[0028] x il Represents the snapshot signal of the i-th frequency at the l-th frame number:

[0029]

[0030] v x (t) = Bs(t) + n x (t);

[0031] E[n p (t)n x (t)]=0 M×M ;

[0032] Where p(t) represents the sound pressure at the location of the vector microphone at time t, n p (t) represents the sound pressure noise at the location of the vector microphone at time t, n x (t) represents the noise at the position of the vector microphone on the x-axis at time t, 0 M×M is an M×M dimensional zero matrix, E represents the desired operation, B is the steering vector, s(t) represents the incident signal vector, H represents the conjugate transpose, and v x (t) represents the vibration velocity of the vector microphone position on the x-axis at time t.

[0033] Furthermore, the objective function f of the MNMF is:

[0034]

[0035] Among them, the elements in the basis matrix T represent the amplitude of the sound source at different frequencies, the elements in the coefficient matrix V represent the gain of the short-time Fourier transform of the sound source at different frame numbers, H is the spatial characteristic matrix, i represents the frequency index, l represents the frame index, tr represents the trace of the matrix, X il represents the spatial correlation matrix, det is the determinant, λ1 and λ2 are the preset regularization parameters, represents the square of the Frobenius norm, I represents the identity matrix, and the superscript T in the formula represents the transpose of the matrix.

[0036] Furthermore, the variables of MNMF are adjusted by the multiplication update method, and the spatial correlation matrix is ​​gradually updated so that the difference between the spatial correlation matrix estimate and the spatial correlation matrix is ​​less than the preset target. The steps of obtaining the optimal spatial correlation matrix estimate include:

[0037] An auxiliary function for constructing the objective function f based on the principalization-minimization algorithm;

[0038] The basis matrix T and the coefficient matrix V are adjusted by the multiplication update method, and the spatial characteristic matrix H is updated based on the adjusted basis matrix T and the adjusted coefficient matrix V until the IS divergence converges, thereby obtaining the optimized basis matrix, coefficient matrix and spatial characteristic matrix;

[0039] The optimal spatial correlation matrix estimate is calculated based on the optimized basis matrix, coefficient matrix and spatial characteristic matrix.

[0040] Furthermore, the step of separating the sound signal based on the optimal spatial correlation matrix includes:

[0041] After the optimal spatial correlation matrix is ​​multiplied by the phase spectrum of the sound signal, an inverse short-time Fourier transform is performed to obtain each separated sound source signal.

[0042] The second aspect of the present invention discloses a sound separation system based on a vector microphone, comprising:

[0043] A receiving module, configured to receive a sound signal including a plurality of incident signals using a vector microphone;

[0044] Angle estimation module, used to construct a fourth-order cumulant matrix of the sound signal, and solve the sound source direction angle estimation and sound source pitch angle estimation of each incident signal based on the fourth-order cumulant matrix;

[0045] A matrix construction module, used to construct a spatial characteristic matrix of each incident signal at different frequencies based on the sound source direction angle estimation and the sound source elevation angle estimation;

[0046] A matrix calculation module is used to calculate the spatial correlation matrix estimation of the sound signal at different frequencies and different frame numbers based on the spatial characteristic matrix, and to construct the spatial correlation matrix of the sound signal at different frequencies and different frame numbers;

[0047] A model building module, used for building an orthogonalized MNMF model based on the spatial characteristic matrix, the spatial correlation matrix estimate and the spatial correlation matrix;

[0048] The optimization module is used to adjust the variables of MNMF by multiplication update method, gradually update the spatial correlation matrix, make the difference between the spatial correlation matrix estimate and the spatial correlation matrix smaller than the preset target, and obtain the optimal spatial correlation matrix estimate;

[0049] The separation module is used to separate the sound signal based on the optimal spatial correlation matrix.

[0050] The sound separation method based on vector microphone proposed in the present invention improves the signal-to-noise ratio by utilizing the data space discrimination and noise suppression characteristics of the high-order covariance matrix, thereby effectively coping with the problem of Gaussian noise; it also constructs a spatial characteristic matrix containing angle information, which can improve the discrimination of vector space information features when the source signal spacing is close, designs a multi-channel non-negative matrix orthogonal constraint cost function, and optimizes the iterative update rules, thereby effectively solving the sound separation problem when the sound source spacing is close and improving the accuracy of sound separation. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. 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 creative work.

[0052] Figure 1 is a flow chart of a sound separation method based on a vector microphone disclosed in an embodiment of the present invention;

[0053] Figure 2 It is a structural diagram of a sound separation system based on a vector microphone disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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.

[0055] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, or product end including a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0056] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0057] Figure 1 A flow chart of a sound separation method based on a vector microphone disclosed in the present invention is shown. The specific steps of the present invention are introduced below in conjunction with a specific embodiment:

[0058] 1. Sound collection

[0059] A vector microphone is a device that can obtain spatial sound information. It receives sound from different directions and generates multi-channel audio signals. These different audio signals can represent the distribution of sound in different directions, thus forming a vector field. These signals usually need to be pre-processed (e.g., filtering, denoising, etc.) to reduce the influence of noise and other irrelevant signals.

[0060] The aperture of the vector microphone is small, which can increase the directional resolution of the system, thereby improving the accuracy of sound source positioning. The vector microphone used in this embodiment consists of a sound pressure microphone and three particle velocity sensors for orthogonal measurement at the same point, which can realize the measurement of single-point sound pressure in any space and particle velocity in three-dimensional orthogonal directions. The signal receiving model of the vector microphone is:

[0061]

[0062] Where x(t) = [p(t),v z (t),v x (t),v y (t)] is the sound pressure and triaxial orthogonal vibration velocity signal, t represents time, p(t) represents the sound pressure at the position of the vector microphone at time t, v x (t), v y (t), v z (t) represents the vibration speed of the vector microphone at the x-axis, y-axis, and z-axis at time t, P represents the number of incident signals, and the incident signal vector s(t) = [s1(t),…,s p (t),…,s P (t)] T ,s p (t) is the pth incident signal, B=[b1,b2,…,b p ,…,b P ] is the steering vector, n(t)=[n p (t),nz (t),n x (t),n y (t)] T 4×L k dimensional noise vector, n p (t) represents the sound pressure noise at the location of the vector microphone at time t, n x (t), n y (t), n z (t) represents the noise of the x-axis, y-axis, and z-axis at the location of the vector microphone at time t, L k The number of snapshots.

[0063] The steering vector b of the pth incident signal p for:

[0064]

[0065] Among them, the azimuth angle θ p ∈[0,2π], pitch angle in are the direction cosines of the x-axis, y-axis and z-axis respectively.

[0066] Define the 4×L received by the vector microphone k The dimensional data is the microphone receiving data x:

[0067]

[0068] In the formula,

[0069] 2. High-order vector space azimuth estimation

[0070] The fourth-order cumulant is solved for the acquired time-domain observed multi-channel data x, and the estimated factorization of the spatial correlation matrix is ​​obtained by introducing the high-order spatial characteristic matrix. After spectral search, the precise azimuth information of the sound source signal can be obtained. The azimuth information of the sound source signal is introduced into the multi-channel non-negative matrix model, which provides prior information for the spatial covariance matrix and improves the sound separation effect.

[0071] The fourth-order cumulant matrix C of the signal receiving model x(t) of the vector microphone x for:

[0072]

[0073] Among them, cum() represents the cumulative amount calculation, x(a,l k ) and x(:,l k ) are the (a,l)th k ) elements and lk List, is the ath element of the steering vector B, s4(l k ) is the lth incident signal of the fourth channel of the vector microphone k elements, * represents complex conjugate, H represents conjugate transpose, n(a,l k ) represents the ath element of the fourth channel in the noise vector, each channel corresponds to a sensor in the vector microphone, and in this embodiment, the vector microphone has four sensors. Smooth the cumulative result to reduce the error. Based on the Gaussian assumption, cum(n(4,l k ),n*(a,l k ),n(:,l k ),n H (:,l k ))=0, then the fourth-order cumulant matrix C x Simplified to:

[0074]

[0075] in, is the fourth-order cumulant of s(t), det(Ψ)≠0, Ψ is of full rank, and the Gaussian measurement noise of each channel in AVS can be suppressed based on high-order cumulants.

[0076] For the fourth-order cumulant matrix C x Perform eigendecomposition and obtain the signal subspace U s,p and the noise subspace U v,p , and based on the noise subspace U v,p Solving the sound source direction angle estimation Sum source pitch angle estimation

[0077]

[0078] 3. Constructing the spatial correlation matrix

[0079] Perform short-time Fourier transform on the acquired time domain observation signal.

[0080] Sound source direction angle estimation based on fourth-order cumulant estimation Sum source pitch angle estimation Construct the spatial characteristics matrix:

[0081] H ip =BB H ;

[0082] B=[b1,b2,…,b p ,…,b P ]

[0083]

[0084] Among them, H ip is the spatial characteristic matrix of the pth incident signal at frequency i, where is the sound source direction angle estimate of the pth incident signal, is the estimated sound source elevation angle of the pth incident signal.

[0085] The vector microphone output snapshot signal is expressed as:

[0086] p(t)=Bs(t)+n p (t);

[0087] v x (t) = Bs(t) + n x (t);

[0088] v y (t) = Bs(t) + n y (t);

[0089] v z (t) = Bs(t) + n z (t);

[0090] v x (t), v y (t), v z (t) represents the vibration speed of the vector microphone at the time t on the x, y, and z axes.

[0091] Construct a covariance matrix containing sound pressure and three-axis particle velocity signals, obtain the sound pressure and x, y, z axis particle velocity signals to construct the spatial covariance matrix pv x , pv y , pv z Assuming that the noise of each channel of the vector sound pressure measurement and the particle velocity measurement is independent, using pv x Combined into the sound pressure velocity covariance matrix, then:

[0092]

[0093] E[n p (t)n x (t)]=0 M×M ;

[0094] Where p(t) represents the sound pressure at the location of the vector microphone at time t, n p (t) represents the sound pressure noise at the location of the vector microphone at time t, n x (t) represents the noise at the location of the vector microphone on the x-axis at time t, x ilrepresents the snapshot signal of the ith frequency at the lth frame number, 0 M×M is an M×M dimensional zero matrix, E represents the expected operation, based on x il Get the spatial correlation matrix X il . It can be expressed as a vector product:

[0095]

[0096] Where i=1,…,I, l=1,…,L are the frequency number and frame number respectively, and H represents the conjugate transpose. The spatial correlation matrix X il The matrix form is as follows:

[0097]

[0098] Among them, the diagonal elements represent the observed signal amplitude |x il |=[|x il1 |,……,|x ilM |] T , the non-diagonal elements represent the phase difference between the rth microphone and the mth microphone.

[0099] Assuming that the sound source position remains unchanged, the spatial characteristic matrix H is introduced ik , which is used to model the spatial characteristics of the kth incident signal at frequency i, where H ik is a Hermitian semi-positive definite matrix with non-negativity. The spatial correlation matrix is ​​modeled in the form of products and obtained as follows:

[0100]

[0101] Among them, t ik is the amplitude of the kth incident signal at frequency i, v kl is the gain of the short-time Fourier transform of the kth incident signal under the frame number l, K represents the number of incident signals, H ik Represents the spatial characteristic matrix of the kth incident signal at frequency i.

[0102] 4. Model Optimization

[0103] The cost function of the orthogonal constrained multi-channel non-negative matrix is ​​designed, orthogonal constraints are imposed on the basis matrix T and the coefficient matrix V, and an auxiliary function is constructed to solve the optimization iterative update rule to obtain the non-negative matrix basis matrix and coefficient matrix of the sound separation algorithm based on the vector microphone.

[0104] The amplitude t of the base k of the sound source at frequency i ik Construct the basis matrix T, with the gain v of the short-time Fourier transform of the sound source in the frame number l with the basis k klConstruct the coefficient matrix V, use the Lagrange multiplier method to construct the MNMF (Multichannel Nonnegative Matrix Factorization), the basis matrix T and coefficient matrix V optimization model in the Multichannel Nonnegative Matrix Factorization (MNMF) model, and define the objective function f as:

[0105]

[0106] Among them, d * Take the IS divergence metric and evaluate X il and The approximation degree between them, det is the determinant, H is the spatial characteristic matrix, λ1 and λ2 are the preset regularization parameters, represents the square of the Frobenius norm and I represents the identity matrix.

[0107] Construct an auxiliary function f of the objective function f based on the Majorization-Minimization (MM) algorithm + :

[0108]

[0109] Where R is the first auxiliary variable matrix, whose elements are R il , U is the second auxiliary variable matrix, where the elements are U il , R il and U il is a positive definite auxiliary variable, and

[0110] Auxiliary function f + satisfy:

[0111] 1)f(T,V,H)≤f + (T, V, H, R, U);

[0112] 2) f(T, V, H) = min R,U f + (T, V, H, R, U);

[0113] When R il and U il When the following equation is satisfied, the objective function f and the auxiliary function f + equivalence:

[0114]

[0115] Based on the multiplication update method, the basis matrix T and the coefficient matrix V are adjusted to minimize the objective function f, where the basis matrix T is updated as follows:

[0116]

[0117] The coefficient matrix V is updated as follows:

[0118]

[0119] Based on the adjusted basis matrix T and coefficient matrix V, the spatial characteristic matrix H is updated by solving the Riccati equation:

[0120] H ik AH ik =Q;

[0121]

[0122] Among them, H i ′ k H before this adjustment ik The corresponding value.

[0123] Adjust the basis matrix T, coefficient matrix V and spatial characteristic matrix H until the IS (Itakura-Saito) divergence converges to obtain the optimized basis matrix T final , the optimized coefficient matrix V final And the optimized spatial characteristic matrix H final .

[0124] 5. Construct the sound source separation signal

[0125] Based on the optimized basis matrix T final , the optimized coefficient matrix V final And the optimized spatial characteristic matrix H final ,according to Calculate the spatial correlation matrix estimate of each separated acoustic signal After multiplying by the phase spectrum of the original signal, an inverse short-time Fourier transform is performed to obtain the separated sound signals.

[0126] Figure 2 The structure diagram of a sound separation system based on a vector microphone disclosed in the present invention is shown, and the sound separation system comprises:

[0127] A receiving module, configured to receive a sound signal including a plurality of incident signals using a vector microphone;

[0128] Angle estimation module, used to construct a fourth-order cumulant matrix of the sound signal, and solve the sound source direction angle estimation and sound source pitch angle estimation of each incident signal based on the fourth-order cumulant matrix;

[0129] A matrix construction module, used to construct a spatial characteristic matrix of each incident signal at different frequencies based on the sound source direction angle estimation and the sound source elevation angle estimation;

[0130] A matrix calculation module is used to calculate the spatial correlation matrix estimation of the sound signal at different frequencies and different frame numbers based on the spatial characteristic matrix, and to construct the spatial correlation matrix of the sound signal at different frequencies and different frame numbers;

[0131] A model building module, used for building an orthogonalized MNMF model based on the spatial characteristic matrix, the spatial correlation matrix estimate and the spatial correlation matrix;

[0132] The optimization module is used to adjust the variables of MNMF by multiplication update method, gradually update the spatial characteristic matrix, make the difference between the spatial correlation matrix estimate and the spatial correlation matrix smaller than the preset target, and obtain the optimal spatial correlation matrix estimate;

[0133] The separation module is used to separate the sound signal based on the optimal spatial correlation matrix.

[0134] In summary, the present invention discloses a sound separation method and system based on a vector microphone, which effectively addresses the problem of Gaussian noise by improving the signal-to-noise ratio through utilizing the data space discrimination and noise suppression characteristics of the high-order covariance matrix; it also constructs a spatial characteristic matrix containing angle information, which can improve the discrimination of vector space information characteristics when the source signal spacing is close, designs a multi-channel non-negative matrix orthogonal constraint cost function, and optimizes the iterative update rule, thereby effectively solving the sound separation problem when the sound source spacing is close, and improving the accuracy of sound separation. Therefore, the present invention effectively overcomes various shortcomings in the prior art and has a high industrial utilization value.

[0135] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A sound separation method based on a vector microphone, characterized in that: include: receiving, using a vector microphone, an acoustic signal comprising a plurality of incident signals; Construct a fourth-order cumulant matrix of the sound signal, and solve the sound source direction angle estimation and sound source pitch angle estimation of each incident signal based on the fourth-order cumulant matrix; Based on the estimation of the sound source direction angle and the sound source elevation angle, the spatial characteristic matrix of each incident signal at different frequencies is constructed; Based on the spatial characteristic matrix, the spatial correlation matrix estimation of the sound signal at different frequencies and different frame numbers is calculated, and the spatial correlation matrix of the sound signal at different frequencies and different frame numbers is constructed; Constructing an orthogonalized MNMF model based on the spatial characteristic matrix, the spatial correlation matrix estimate and the spatial correlation matrix; The variables of MNMF are adjusted by the multiplication update method, and the spatial characteristic matrix is ​​gradually updated so that the difference between the spatial correlation matrix estimate and the spatial correlation matrix is ​​less than the preset target, and the optimal spatial correlation matrix estimate is obtained; Separate the sound signals based on the optimal spatial correlation matrix.

2. The sound separation method based on vector microphone according to claim 1, characterized in that: The sound signal is the sum of the target signal and the noise. The target signal is the sum of the products of each incident signal and its guide vector. The noise includes sound pressure noise and vibration noise on the x, y and z axes.

3. The sound separation method based on vector microphone according to claim 1, characterized in that: The steps of solving the sound source direction angle estimation and sound source elevation angle estimation of each incident signal based on the fourth-order cumulant matrix include: Perform eigendecomposition on the fourth-order cumulant matrix to obtain the signal subspace and noise subspace of the sound signal; Calculate multiple sound source direction angles and multiple sound source elevation angles of each incident signal based on the noise subspace; According to the correlation between the signal subspace and the noise subspace, the sound source direction angle estimation and the sound source pitch angle estimation of the corresponding incident signal are screened out from multiple sound source direction angles and pitch angles.

4. The sound separation method based on vector microphone according to claim 1, characterized in that: Construct the spatial characteristic matrix of each incident signal at different frequencies. For the spatial characteristic matrix H of the pth incident signal at frequency i iP for: in, is the sound source direction angle estimate of the pth incident signal, is the estimated sound source elevation angle of the pth incident signal.

5. The sound separation method based on vector microphone according to claim 4 is characterized in that: Based on the spatial characteristic matrix, the steps of calculating the spatial correlation matrix estimation of the sound signal at different frequencies and different frame numbers include: Spatial correlation matrix estimation at the i-th frequency and frame number l Among them, t ik is the amplitude of the kth incident signal at frequency i, v kl is the gain of the short-time Fourier transform of the kth incident signal under the frame number l, K represents the number of incident signals, H ik Represents the spatial characteristic matrix of the kth incident signal at frequency i.

6. The sound separation method based on vector microphone according to claim 5, characterized in that: The steps of constructing the spatial correlation matrix of the sound signal at different frequencies and different frame numbers include: The spatial correlation matrix X at the i-th frequency and frame number l il : x il Represents the snapshot signal of the i-th frequency at the l-th frame number: v x (t)=Bs(t)+n x (t); E[n p (t)n x (t)]=0 M×M ; Where p(t) represents the sound pressure at the location of the vector microphone at time t, n p (t) represents the sound pressure noise at the location of the vector microphone at time t, n x (t) represents the noise at the position of the vector microphone on the x-axis at time t, 0 M×M is an M×M dimensional zero matrix, E represents the desired operation, B is the steering vector, s(t) represents the incident signal vector, H represents the conjugate transpose, and v x (t) represents the vibration velocity of the vector microphone position on the x-axis at time t.

7. The sound separation method based on vector microphone according to claim 6, characterized in that: The objective function f of the MNMF is: Among them, the elements in the basis matrix T represent the amplitude of the sound source at different frequencies, the elements in the coefficient matrix V represent the gain of the short-time Fourier transform of the sound source at different frame numbers, H is the spatial characteristic matrix, i represents the frequency index, l represents the frame index, tr represents the trace of the matrix, X il represents the spatial correlation matrix, det is the determinant, λ1 and λ2 are the preset regularization parameters, represents the square of the Frobenius norm, I represents the identity matrix, and the superscript T in the formula represents the transpose of the matrix.

8. The sound separation method based on vector microphone according to claim 7, characterized in that: The variables of MNMF are adjusted by the multiplication update method, and the spatial correlation matrix is ​​gradually updated so that the difference between the spatial correlation matrix estimate and the spatial correlation matrix is ​​less than the preset target. The steps of obtaining the optimal spatial correlation matrix estimate include: An auxiliary function for constructing the objective function f based on the principalization-minimization algorithm; The basis matrix T and the coefficient matrix V are adjusted by the multiplication update method, and the spatial characteristic matrix H is updated based on the adjusted basis matrix T and the adjusted coefficient matrix V until the IS divergence converges, thereby obtaining the optimized basis matrix, coefficient matrix and spatial characteristic matrix; The optimal spatial correlation matrix estimate is calculated based on the optimized basis matrix, coefficient matrix and spatial characteristic matrix.

9. The sound separation method based on vector microphone according to claim 1, characterized in that: The steps of separating the sound signal based on the optimal spatial correlation matrix include: After the optimal spatial correlation matrix is ​​multiplied by the phase spectrum of the sound signal, an inverse short-time Fourier transform is performed to obtain each separated sound source signal.

10. A sound separation system based on a vector microphone, characterized in that: include: A receiving module, configured to receive a sound signal including a plurality of incident signals using a vector microphone; Angle estimation module, used to construct a fourth-order cumulant matrix of the sound signal, and solve the sound source direction angle estimation and sound source pitch angle estimation of each incident signal based on the fourth-order cumulant matrix; A matrix construction module, used to construct a spatial characteristic matrix of each incident signal at different frequencies based on the sound source direction angle estimation and the sound source elevation angle estimation; A matrix calculation module is used to calculate the spatial correlation matrix estimation of the sound signal at different frequencies and different frame numbers based on the spatial characteristic matrix, and to construct the spatial correlation matrix of the sound signal at different frequencies and different frame numbers; A model building module, used for building an orthogonalized MNMF model based on the spatial characteristic matrix, the spatial correlation matrix estimate and the spatial correlation matrix; The optimization module is used to adjust the variables of MNMF by multiplication update method, gradually update the spatial correlation matrix, make the difference between the spatial correlation matrix estimate and the spatial correlation matrix smaller than the preset target, and obtain the optimal spatial correlation matrix estimate; The separation module is used to separate the sound signal based on the optimal spatial correlation matrix.

Images