A sound source localization method and system based on polyhedral microphone array

Abstract

The present invention provides a sound source localization method based on a polyhedral microphone array, comprising: S1: collecting multi-channel data collected by multiple microphones forming a polyhedral structure; S2: iteratively processing the multi-channel data using an independent component analysis algorithm to obtain noise-reduced target sound source data; and S3: determining the orientation of the target sound source based on the attenuation of the target sound source by each microphone. The advantages of the present invention are that by collecting sound data using a polyhedral microphone array, the microphone array receives signals that form a non-singular matrix, thereby enabling the extraction of the sound source using an independent component analysis algorithm, and then determining its orientation relative to the microphone array based on the attenuation of the sound source. This further achieves sound source localization based on sound source separation, and can meet the needs of scenarios such as abnormal sound source localization and fault detection.

Description

Technical Field

[0001] The present invention relates to the technical field of sound localization, and in particular to a sound localization method and system based on a polyhedral microphone array. Background Art

[0002] Independent Component Analysis (ICA), as a statistical signal data analysis tool, can effectively blindly separate independent source signals from complex linear mixtures without prior knowledge. This technology has been widely used in fields such as vibration fault monitoring and image feature extraction. Li Ru's invention patent application with publication number CN109409341A discloses a method for identifying approaching aircraft engine noise sources based on the ICA model.

[0003] A prerequisite for the ICA algorithm to effectively separate independent source signals is that the array's received signals are linearly mixed, and the rank of the mixing matrix is ​​no less than the number of signal sources. In vibration detection applications, the received signals from the sensor array typically meet these conditions. Different distances between the sensor and the vibration source will result in different attenuations. Furthermore, because sound waves propagate rapidly in solids (for example, at 5000 m / s in iron), the delay differences caused by varying sensor distances can be ignored within a certain range. Even with some distance differences between the sensor and the sound source, a linear superposition model can still be approximated. However, detection in air is different, where the speed of sound propagates at approximately 340 m / s, and delay errors cannot be ignored. If the microphone array is placed very close together to minimize delay errors, the attenuations of the individual microphones become too similar, and the mixing matrix approaches a singular matrix, resulting in numerical instability in the ICA calculation and unreliable results. Therefore, linear ICA is not ideal for detecting sound signals propagating in air. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method for sound detection and positioning in an air environment using ICA by using a polyhedral microphone array.

[0005] The present invention solves the above technical problems through the following technical solutions: a sound source localization method based on a polyhedral microphone array, comprising:

[0006] S1: collects multi-channel data collected by multiple microphones forming a polyhedron structure;

[0007] The polyhedral microphone array includes a plurality of non-omnidirectional microphones whose normal directions are not parallel, and the distance between adjacent microphones is less than 1 / 5 of the wavelength of the target acoustic signal;

[0008] S2: Iteratively process multi-channel data through the independent component analysis algorithm to obtain the target sound source data after noise reduction;

[0009] S3: Determine the direction of the target sound source based on the attenuation of the target sound source by each microphone.

[0010] The present invention collects sound data through a polyhedral microphone array, so that the signal received by the microphone array approximately satisfies the linear superposition model, and the mixing matrix is ​​a non-singular matrix, so that the sound source can be extracted through the independent component analysis algorithm, and then its orientation relative to the microphone array is judged by the attenuation of the sound source. On the basis of sound source separation, the sound source positioning is further realized, which can meet the needs of scenarios such as abnormal sound source positioning and fault detection.

[0011] Preferably, the method of iteratively processing multi-channel data by the independent component analysis algorithm is:

[0012] Step i: Randomly initialize the unmixing matrix w1 with a modulus of 1, the zero-mean Gaussian vector v with a variance of 1, the whitening matrix Q, the Lagrange multiplier μ1 and the similarity threshold ε, and the number of iterations k = 1;

[0013] Step ii: Perform whitening operation on the received multi-channel data matrix y, that is,

[0014]

[0015] Step iii: Calculate the sound source information x after the kth iteration noise reduction process k ,

[0016]

[0017] Step iv: Calculate x k The similarity with the reference signal r,

[0018] s k =G(x k ,r)-ε

[0019] The formula of the similarity function is,

[0020] G(x k ,r)=-E{|x k 2 |·r}

[0021] E{} represents the expected value;

[0022] Step v: Update the parameter unmixing matrix and Lagrange multipliers,

[0023] μ k+1 =max{0,μ k +s k}

[0024]

[0025] in,

[0026]

[0027] L(x k ,y,w k )=ρ·E{yF′(x k )}-0.5μ k E{yG′(x k ,r)}

[0028] δ(w k )=ρ·E{yF″(x k )}-0.5μ k E{yG″(x k ,r)}

[0029] ρ=E{F(x)}-E{F(v)}

[0030]

[0031] η is a constant, F′(x k ) and F″(x k ) respectively represent F(x k )’s first and second derivatives;

[0032] Step vi: If Δ is the preset empirical value, let k = k + 1, return to step iii, otherwise output x k .

[0033] Preferably, the method for determining the direction of the target sound source is:

[0034] The mixing matrix of the desired sound source obtained by the unmixing matrix and the whitening matrix is:

[0035] W mix =pinv(w k ,Q)

[0036] Among them, pinv() represents the generalized inverse matrix, W mix is a vector of attenuation coefficients from the sound source to each microphone,

[0037] According to the position and direction of each microphone, the attenuation coefficient vector A of the sound in each direction relative to the polyhedral microphone is calculated. n ,n=1,2,···,N, get the attenuation coefficient vectors in N directions, the sound source direction is,

[0038]

[0039] The corresponding direction is the estimated direction of the desired sound source.

[0040] Preferably, the shape of the polyhedral microphone array includes a first box in the center and four second boxes connected to the four sides of the first box respectively, the normal angles between the four second boxes and the first box are less than 90° and equal, and a microphone is respectively provided on the first box and each second box.

[0041] The present invention also provides a sound source localization system based on a polyhedral microphone array, comprising:

[0042] A data acquisition module, used for acquiring multi-channel data collected by multiple microphones forming a polyhedron structure;

[0043] The polyhedral microphone array includes a plurality of non-omnidirectional microphones whose normal directions are not parallel, and the distance between adjacent microphones is less than 1 / 5 of the wavelength of the desired acoustic signal;

[0044] The noise reduction module iteratively processes multi-channel data through the independent component analysis algorithm to obtain the target sound source data after noise reduction;

[0045] The positioning module determines the direction of the target sound source based on the attenuation of the target sound source by each microphone.

[0046] Preferably, the method of iteratively processing multi-channel data by the independent component analysis algorithm is:

[0047] Step i: Randomly initialize the unmixing matrix w1 with a modulus of 1, the zero-mean Gaussian vector v with a variance of 1, the whitening matrix Q, the Lagrange multiplier μ1 and the similarity threshold ε, and the number of iterations k = 1;

[0048] Step ii: Perform whitening operation on the received multi-channel data matrix y, that is,

[0049]

[0050] Step iii: Calculate the sound source information x after the kth iteration noise reduction process k ,

[0051]

[0052] Step iv: Calculate x k The similarity with the reference signal r,

[0053] s k =G(x k ,r)-ε

[0054] The formula of the similarity function is,

[0055] G(x k ,r)=-E{|x k 2 |·r}

[0056] E{} represents the expected value;

[0057] Step v: Update the parameter unmixing matrix and Lagrange multipliers,

[0058] μ k+1 =max{0,μ k +s k}

[0059]

[0060] in,

[0061]

[0062] L(x k ,y,w k )=ρ·E{yF′(x k )}-0.5μ k E{yG′(x k ,r)}

[0063] δ(w k )=ρ·E{yF″(x k )}-0.5μ k E{yG″(x k ,r)}

[0064] ρ=E{F(x)}-E{F(v)}

[0065]

[0066] η is a constant, F′(x k ) and F″(x k ) respectively represent F(x k )’s first and second derivatives;

[0067] Step vi: If Δ is the preset empirical value, let k = k + 1, return to step iii, otherwise output x k .

[0068] Preferably, the method for determining the direction of the target sound source is:

[0069] The mixing matrix of the desired sound source obtained by the unmixing matrix and the whitening matrix is:

[0070] W mix =pinv(w k ,Q)

[0071] Among them, pinv() represents the generalized inverse matrix, W mix is a vector of attenuation coefficients from the sound source to each microphone,

[0072] According to the position and direction of each microphone, the attenuation coefficient vector A of the sound in each direction relative to the polyhedral microphone is calculated. n ,n=1,2,···,N, get the attenuation coefficient vectors in N directions, the sound source direction is,

[0073]

[0074] The corresponding direction is the estimated direction of the desired sound source.

[0075] Preferably, the shape of the polyhedral microphone array includes a first box in the center and four second boxes connected to the four sides of the first box respectively, the normal angles between the four second boxes and the first box are less than 90° and equal, and a microphone is respectively provided on the first box and each second box.

[0076] The present invention also provides an electronic processing device, comprising at least one processor and a storage device storing at least one execution program. When the at least one execution program is executed by the at least one processor, the at least one processor implements the sound source localization method as described above.

[0077] The present invention also provides a computer-readable storage medium storing a computer program, wherein the computer program can implement the sound source localization method when executed by a processor.

[0078] The advantages of the sound source localization method based on a polyhedral microphone array provided by the present invention are: sound data is collected by the polyhedral microphone array, so that the signal received by the microphone array approximately constitutes a linear superposition model, and the mixing matrix is ​​a non-singular matrix, so that the sound source can be extracted through the independent component analysis algorithm, and then its orientation relative to the microphone array is judged by the attenuation of the sound source. On the basis of sound source separation, the sound source positioning is further realized, which can meet the needs of scenarios such as abnormal sound source positioning and fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 A flowchart of a sound source localization method based on a polyhedral microphone array provided by an embodiment of the present invention;

[0080] Figure 2 A schematic diagram of the structure of a polyhedral microphone array for a sound source localization method based on a polyhedral microphone array provided in an embodiment of the present invention;

[0081] Figure 3 A schematic diagram of signal attenuation of a non-omnidirectional microphone provided by an embodiment of the present invention;

[0082] Figure 4 A schematic diagram of two non-omnidirectional microphone array states provided by an embodiment of the present invention;

[0083] Figure 5 A schematic diagram of a sound source separation circuit for a sound source localization method based on a polyhedral microphone array provided by an embodiment of the present invention;

[0084] Figure 6 A schematic diagram of a sound source localization system based on a polyhedral microphone array provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0085] To make the objectives, technical solutions, and advantages of the present invention more clearly understood, the following describes the technical solutions of the present invention clearly and completely with reference to specific embodiments and the accompanying drawings. It is obvious that the embodiments described are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0086] like Figure 1 As shown, this embodiment provides a sound source localization method based on a polyhedral microphone array, comprising:

[0087] S1: collects multi-channel data collected by multiple microphones forming a polyhedron structure;

[0088] The polyhedral microphone array includes a plurality of non-omnidirectional microphones whose normal directions are not parallel, and the distance between adjacent microphones is less than the wavelength of the target acoustic signal;

[0089] S2: Iteratively process multi-channel data through the independent component analysis algorithm to obtain the target sound source data after noise reduction;

[0090] S3: Determine the direction of the target sound source based on the attenuation of the target sound source by each microphone.

[0091] This embodiment collects sound data through a polyhedral microphone array, so that the signals received by the microphone array approximately constitute a linear superposition model, and the mixing matrix is ​​a non-singular matrix, so that the sound source can be extracted through the independent component analysis algorithm, and then its position relative to the microphone array is judged by the attenuation of the sound source. On the basis of sound source separation, the sound source positioning is further realized, which can meet the needs of scenarios such as abnormal sound source positioning and fault detection.

[0092] Specifically, the sound source localization method based on the polyhedral microphone array provided in this embodiment includes:

[0093] S1: Collect multi-channel data collected by multiple microphones that form a polyhedral structure.

[0094] The polyhedral microphone array includes multiple non-omnidirectional microphones with non-parallel normal directions. The distance between adjacent microphones is less than the wavelength of the target acoustic signal, preferably less than 1 / 5 of the wavelength of the target acoustic signal. On the basis of meeting the above conditions, the polyhedral microphone array can be set to any structure.

[0095] Reference Figure 2 , in this embodiment, the polyhedral microphone array structure includes a first square frame at the center and second square frames respectively connected to the four sides of the first square frame. The normal angles between the four second square frames and the first square frame are equal and less than 90°. One microphone is provided on each of the first square frame and the second square frames, so as to meet the requirements of non-parallel normal directions and the distance being less than the wavelength of the acoustic signal. The microphone is selected as a non-omnidirectional microphone. Combining Figure 3 , the non-omnidirectional microphone has different sound attenuations in different directions, with the smallest attenuation in the normal direction. The larger the angle from the normal direction, the greater the attenuation. Combining Figure 4 , for an array composed of two non-omnidirectional microphones, the two microphones are in different planes and are both non-omnidirectional microphones. The interfering sound source s1 and the sound source s2 to be detected radiate sound waves to the microphone array from different positions respectively. Let the attenuation coefficients of the interfering sound source s1 to the two microphones be a1 and a2 respectively, and the attenuation coefficients of the sound source s2 to be detected to the two microphones be b1 and b2 respectively. Since the array spacing is very small, much smaller than the signal wavelength, the time delay can be ignored here, and only the amplitude attenuation is considered. Then the amplitude model of the array received signal is

[0096]

[0097] In matrix form,

[0098]

[0099] where A is the mixing matrix. An important condition for the independent component analysis algorithm to effectively separate the signal s2 to be detected from the mixed signal is that the mixing matrix A is a non-singular matrix. According to Figure 3 and Figure 4 it can be seen that due to the different directions of the two microphones, there must be a1 > a2 and b1 < b2, so the mixing matrix A must be a non-singular matrix.

[0100] S2: Iteratively process the multi-channel data through the independent component analysis algorithm to obtain the target sound source data after noise reduction.

[0101] Combining Figure 5 , the method for the independent component analysis algorithm to iteratively process the multi-channel data is

[0102] Step i: Randomly initialize the unmixing matrix w1 with a modulus of 1, the zero-mean Gaussian vector v with a variance of 1, the whitening matrix Q, the Lagrange multiplier μ1 and the similarity threshold ε, and the number of iterations k = 1;

[0103] Step ii: Perform whitening operation on the received multi-channel data matrix y, that is,

[0104]

[0105] Step iii: Calculate the sound source information x after the kth iteration noise reduction process k ,

[0106]

[0107] Step iv: Calculate x k The similarity with the reference signal r,

[0108] s k =G(x k ,r)-ε

[0109] The formula of the similarity function is,

[0110] G(x k ,r)=-E{|x k 2 |·r}

[0111] E{} represents the expected value;

[0112] Step v: Update the parameter unmixing matrix and Lagrange multipliers,

[0113] μ k+1 =max{0,μ k +s k}

[0114]

[0115] in,

[0116]

[0117] L(x k ,y,w k )=ρ·E{yF′(x k )}-0.5μ k E{yG′(x k ,r)}

[0118] δ(w k )=ρ·E{yF″(x k )}-0.5μ kE{yG″(x k ,r)}

[0119] ρ=E{F(x)}-E{F(v)}

[0120]

[0121] η is a constant, which is an empirical value, F′(x k ) and F″(x k ) respectively represent F(x k )’s first and second derivatives;

[0122] Step vi: If Δ is the preset empirical value, let k = k + 1, return to step iii, otherwise output x k .

[0123] S3: Determine the direction of the target sound source based on the attenuation of the target sound source by each microphone, using the following method:

[0124] The mixing matrix of the desired sound source obtained by the unmixing matrix and the whitening matrix is:

[0125] W mix =pinv(w k ,Q)

[0126] Among them, pinv() represents the generalized inverse matrix, W mix is a vector of attenuation coefficients from the sound source to each microphone,

[0127] According to the position and direction of each microphone, the attenuation coefficient vector A of the sound in each direction relative to the polyhedral microphone is calculated. n ,n=1,2,···,N, get the attenuation coefficient vectors in N directions, the sound source direction is,

[0128]

[0129] The corresponding direction is the estimated direction of the desired sound source.

[0130] When constructing a polyhedral microphone, the attenuation coefficient vector A can be calculated or experimentally determined. n , and then the estimated direction of the sound source can be determined by calculation to achieve sound source positioning.

[0131] Combine Figure 6 This embodiment also provides a sound source localization system based on a polyhedral microphone array, comprising:

[0132] A data acquisition module, used for acquiring multi-channel data collected by multiple microphones forming a polyhedron structure;

[0133] The polyhedral microphone array includes a plurality of non-omnidirectional microphones whose normal directions are not parallel, and the distance between adjacent microphones is less than 1 / 5 of the wavelength of the target acoustic signal;

[0134] The noise reduction module iteratively processes multi-channel data through the independent component analysis algorithm to obtain the target sound source data after noise reduction;

[0135] The positioning module determines the direction of the target sound source based on the attenuation of the target sound source by each microphone.

[0136] This embodiment further provides an electronic processing device, comprising at least one processor and a storage device storing at least one execution program. When the at least one execution program is executed by the at least one processor, the at least one processor performs the following method:

[0137] S1: collects multi-channel data collected by multiple microphones forming a polyhedron structure;

[0138] The polyhedral microphone array includes a plurality of non-omnidirectional microphones whose normal directions are not parallel, and the distance between adjacent microphones is less than 1 / 5 of the wavelength of the target acoustic signal;

[0139] S2: Iteratively process multi-channel data through the independent component analysis algorithm to obtain the target sound source data after noise reduction;

[0140] S3: Determine the direction of the target sound source based on the attenuation of the target sound source by each microphone.

[0141] This embodiment further provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can implement the following method:

[0142] S1: collects multi-channel data collected by multiple microphones forming a polyhedron structure;

[0143] The polyhedral microphone array includes a plurality of non-omnidirectional microphones whose normal directions are not parallel, and the distance between adjacent microphones is less than 1 / 5 of the wavelength of the target acoustic signal;

[0144] S2: Iteratively process multi-channel data through the independent component analysis algorithm to obtain the target sound source data after noise reduction;

[0145] S3: Determine the direction of the target sound source based on the attenuation of the target sound source by each microphone.

[0146] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A sound source localization method based on a polyhedral microphone array, characterized by: include, S1: collects multi-channel data collected by multiple microphones forming a polyhedron structure; The polyhedral microphone array includes a plurality of non-omnidirectional microphones whose normal directions are not parallel, and the spacing between adjacent microphones is less than 1 / 5 of the wavelength of the target acoustic signal; the shape of the polyhedral microphone array includes a first box in the center and four second boxes connected to the four sides of the first box respectively, the four second boxes have normal angles less than 90° with the first box and are equal, and a microphone is respectively provided on the first box and each second box; S2: Iteratively process multi-channel data through the independent component analysis algorithm to obtain the target sound source data after noise reduction; S3: Determine the direction of the target sound source based on the attenuation of the target sound source by each microphone.

2. The method for sound source localization based on a polyhedral microphone array according to claim 1, wherein: The method of iteratively processing multi-channel data by the independent component analysis algorithm is: Step i: Randomly initialize the unmixing matrix w1 with a modulus of 1, the zero-mean Gaussian vector v with a variance of 1, the whitening matrix Q, the Lagrange multiplier μ1 and the similarity threshold ε, and the number of iterations k = 1; Step ii: Perform whitening operation on the received multi-channel data matrix y, that is, Step iii: Calculate the sound source information x after the k-th iteration noise reduction process k , Step iv: Calculate x k The similarity with the reference signal r, s k =G(x k ,r)-e The formula of the similarity function is, G(x k ,r)=-E{|x k 2 |·r} E{} represents the expected value; Step v: Update the parameter unmixing matrix and Lagrange multipliers, m k+1 =max{0,μ k +s k } in, L(x k ,y,w k )=ρ·E{yF′(x k )}-0.5μ k E{yG′(x k ,r)} δ(w k )=ρ·E{yF″(x k )}-0.5m k E{yG″(x k ,r)} ρ=E{F(x)}-E{F(v)} η is a constant, F′(x k ) and F″(x k ) respectively represent F(x k )’s first and second derivatives; Step vi: If Δ is the preset empirical value, let k = k + 1, return to step iii, otherwise output x k .

3. The sound source localization method based on a polyhedral microphone array according to claim 2, characterized in that: The method for determining the direction of the target sound source is: The mixing matrix of the desired sound source obtained by the unmixing matrix and the whitening matrix is: W mix =pinv(w k ,Q) Among them, pinv() represents the generalized inverse matrix, W mix is a vector of attenuation coefficients from the sound source to each microphone, According to the position and direction of each microphone, the attenuation coefficient vector A of the sound in each direction relative to the polyhedral microphone is calculated. n ,n=1,2,···,N, get the attenuation coefficient vectors in N directions, the sound source direction is, The corresponding direction is the estimated direction of the desired sound source.

4. A sound source localization system based on a polyhedral microphone array, characterized by: include, A data acquisition module, used for acquiring multi-channel data collected by multiple microphones forming a polyhedron structure; The polyhedral microphone array includes a plurality of non-omnidirectional microphones whose normal directions are not parallel, and the spacing between adjacent microphones is less than 1 / 5 of the wavelength of the target acoustic signal; the shape of the polyhedral microphone array includes a first box in the center and four second boxes connected to the four sides of the first box respectively, the four second boxes have normal angles less than 90° with the first box and are equal, and a microphone is respectively provided on the first box and each second box; The noise reduction module iteratively processes multi-channel data through the independent component analysis algorithm to obtain the target sound source data after noise reduction; The positioning module determines the direction of the target sound source based on the attenuation of the target sound source by each microphone.

5. The sound source localization system based on a polyhedral microphone array according to claim 4, characterized in that: The method of iteratively processing multi-channel data by the independent component analysis algorithm is: Step i: Randomly initialize the unmixing matrix w1 with a modulus of 1, the zero-mean Gaussian vector v with a variance of 1, the whitening matrix Q, the Lagrange multiplier μ1 and the similarity threshold ε, and the number of iterations k = 1; Step ii: Perform whitening operation on the received multi-channel data matrix y, that is, Step iii: Calculate the sound source information x after the k-th iteration noise reduction process k , Step iv: Calculate x k The similarity with the reference signal r, s k =G(x k ,r)-e The formula of the similarity function is, G(x k ,r)=-E{|x k 2 |·r} E{} represents the expected value; Step v: Update the parameter unmixing matrix and Lagrange multipliers, m k+1 =max{0,μ k +s k } in, L(x k ,y,w k )=ρ·E{yF′(x k )}-0.5μ k E{yG′(x k ,r)} δ(w k )=ρ·E{yF″(x k )}-0.5m k E{yG″(x k ,r)} ρ=E{F(x)}-E{F(v)} η is a constant, F′(x k ) and F″(x k ) respectively represent F(x k )’s first and second derivatives; Step vi: If Δ is the preset empirical value, let k = k + 1, return to step iii, otherwise output x k .

6. The sound source localization system based on a polyhedral microphone array according to claim 5, characterized in that: The method for determining the direction of the target sound source is: The mixing matrix of the desired sound source obtained by the unmixing matrix and the whitening matrix is: W mix =pinv(w k ,Q) Among them, pinv() represents the generalized inverse matrix, W mix is a vector of attenuation coefficients from the sound source to each microphone, According to the position and direction of each microphone, the attenuation coefficient vector A of the sound in each direction relative to the polyhedral microphone is calculated. n ,n=1,2,···,N, get the attenuation coefficient vectors in N directions, the sound source direction is, The corresponding direction is the estimated direction of the desired sound source.

7. An electronic processing device, characterized in that: The system comprises at least one processor and a storage device storing at least one execution program. When the at least one execution program is executed by the at least one processor, the at least one processor implements the method according to any one of claims 1 to 3.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 3 can be implemented.

Images