Three-dimensional MUSIC sound source localization method based on improved particle swarm optimization algorithm

Abstract

The invention relates to a three-dimensional MUSIC sound source localization method based on an improved particle swarm optimization algorithm, and belongs to the technical field of sound source localization, and the method comprises the following steps: S1, building a covariance matrix for sound signal data collected by an array of M array elements, and separating a noise subspace; s2, performing array manifold construction of a three-dimensional space by adopting a large-step azimuth angle, a large-step pitch angle and a large-step distance; s3, performing preliminary positioning on the sound source according to a MUSIC algorithm principle; and S4, initializing parameters of the particle swarm optimization algorithm for alternate search, and searching a three-dimensional space.

Description

Technical Field

[0001] The present invention belongs to the technical field of sound source localization, and relates to a three-dimensional MUSIC sound source localization method based on an improved particle swarm optimization algorithm. Background Art

[0002] To improve the localization efficiency of three-dimensional MUSIC sound source localization technology, many people adopt the three-dimensional MUSIC sound source localization technology based on high-order cumulants. Although high-order cumulants (High-Order Cumulants, HOC) can decouple angles and distances, improve calculation efficiency, and suppress Gaussian white noise, thereby improving the robustness of the MUSIC algorithm, the positioning performance of this method is low, and both the positioning accuracy rate and the positioning precision are low. When solving the MUSIC algorithm based on high-order cumulants, the root-MUSIC algorithm is often used, but this algorithm is generally only applied to linear arrays. When performing three-dimensional sound source localization in space, it is often necessary to solve the geometric relationship between arrays, introducing a parameter matching problem. At the same time, when the number of microphones is large and the collected data is large, the construction of high-order cumulants will also consume a large amount of computing resources, reducing the calculation efficiency of the algorithm. Although many people have introduced new eigenvalue solutions and high-order cumulant solutions to the MUSIC sound source localization technology based on high-order cumulants, their solution steps require the array to satisfy a certain shape. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide a three-dimensional MUSIC sound source localization method based on an improved particle swarm optimization algorithm.

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

[0005] A three-dimensional MUSIC sound source localization method based on an improved particle swarm optimization algorithm includes the following steps:

[0006] S1: Construct a covariance matrix from the acoustic signal data collected by an array of M array elements and separate the noise subspace;

[0007] S2: Construct the array manifold in three-dimensional space using large step sizes for azimuth, elevation, and distance;

[0008] S3: Perform preliminary localization of the sound source according to the MUSIC algorithm principle;

[0009] S4: Initialize the parameters of the particle swarm optimization algorithm for alternating search and search in three-dimensional space.

[0010] Further, the step S1 includes the following steps:

[0011] S11: Construct a covariance matrix from the acoustic signal data collected by the array elements:

[0012] R = E[y(t)y(t) H

[0013] where R is the constructed covariance matrix, E represents taking the expectation, [·] H represents the transpose, and y(t) is the acoustic signal collected by the array;

[0014] S12: Perform eigenvalue decomposition on the covariance matrix. According to the number of sound sources, take all columns with a column number greater than the number of sound sources as the noise subspace:

[0015]

[0016] where U S represents the signal subspace, Σ S represents the eigenvalues corresponding to the signal subspace, U N represents the noise subspace, and Σ N represents the eigenvalues corresponding to the noise subspace.

[0017] Furthermore, the step S2 includes the following steps:

[0018] Construct the array manifold according to the near-field propagation model of the acoustic signal:

[0019] y(t) = A(f, Φ)s(t) + W(t)

[0020]

[0021] where A(f, Φ) = [a1(f, Φ1), a2(f, Φ2),..., a j (f, Φ N )], a j (f, Φ j ) represents the array manifold, f represents the frequency of the acoustic signal; θ, r represent the azimuth angle, elevation angle, and distance respectively; s(t) represents the amplitude of the sound source, W(t) represents the noise, Γ ji represents the distance between the jth sound source and the ith microphone, and c represents the speed of sound.

[0022] Furthermore, the S3 includes the following steps:

[0023] Perform preliminary positioning using the orthogonality relationship between the array manifold and the noise subspace in the MUSIC algorithm principle to obtain the preliminary positioning result

[0024] P(f, Φ) = ||A(f, Φ) H U N || 2 ​

[0025] Where P(f, Φ) is the reciprocal of the spatial spectrum, and A(f, Φ) H is the constructed array manifold.

[0026] Furthermore, the S4 includes the following steps:

[0027] S41: Initialize the parameters of the particle swarm optimization algorithm, including: the maximum value K of the constraint factor with linear variation dmax and the minimum value K dmin , the maximum number of iterations iter for a single time max , the maximum number of alternations change max , the maximum value ω of the inertial weight with linear variation max and the minimum value ω min , the changing self-learning factor c 1max , c 1middle , c 1min , the boundary ratios λ1, λ2 for controlling the change of the self-learning sub-factor, and the condition γ for stopping iteration; Initialize the parameters to control the parameter change of the particle swarm optimization algorithm during iteration:

[0028] K d = K dmax -(K dmax - K dmin )·u / U

[0029] U = iter max ·change max

[0030] ω = ω max -(ω max - ω min )·u / U

[0031]

[0032] S42: Initialize the velocities of all particles:

[0033]

[0034] Wherein, is the velocity of the k-th particle at the u-th iteration in the d-th dimension, K d is the constraint factor, ω is the inertial weight, indicating the influence of the velocity update of the particle at the previous iteration on the velocity update of the particle at the next iteration, rand3 is a random number between [0, 1], and Value g is the overall optimal fitness value of the particle, which is P(f, Φ) at the initial positioning when searching for the first time;

[0035] S43: First, for the angle θ, Perform traversal and then traverse the distance r:

[0036]

[0037] In the formula, c1 and c2 are the self-learning factor and social learning factor of the particle respectively, representing the influence of the particle's own historical best position and the historical best position of the entire particle swarm on the particle. rand1 and rand2 are random numbers between [0, 1], and pBest k is the historical best fitness function value of the k-th particle itself, and gBest is the historical best fitness function value of the entire particle swarm. is the position of the k-th particle at the u-th iteration in the d-th dimension;

[0038] S44: If gBest < γ or the maximum number of iterations is reached, output the positioning result;

[0039] S45: If there is no change in gBest in U consecutive traversals during a search, swap the search dimension and return to S42; otherwise, continue the traversal in the current dimension.

[0040] The beneficial effects of the present invention are as follows: Since the present invention performs preliminary positioning, the search range can be initially locked to prevent the particle swarm optimization algorithm from falling into a local optimal solution; after observing the spatial spectrum, the present invention adopts an alternating search strategy of angle and distance, which can prevent the blind search of particles and greatly improve the search efficiency. When performing alternating search, the present invention introduces the global optimal fitness function value to initialize the particle position, making the particle distribution more reasonable and improving the search efficiency; the present invention introduces a linearly varying constraint factor to perform different constraints on angle search and distance search, preventing the search from being too divergent or concentrated due to the change in the order of magnitude of the search range; the present invention introduces a linearly transformed inertia weight and self-learning factor to improve the global search ability in the early stage and local search ability in the later stage of the particle swarm optimization algorithm. This method introduces the particle swarm optimization algorithm into the three-dimensional MUSIC sound source localization technology and improves the search strategy, making the calculation efficiency of the three-dimensional MUSIC sound source localization technology improved, and at the same time ensuring the positioning accuracy under high signal-to-noise ratio.

[0041] Since the initial positioning is carried out in the present invention, the search range can be initially locked, preventing the particle swarm optimization algorithm from falling into a local optimal solution; after observing the spatial spectrum in the present invention, an alternating search strategy for angle and distance is adopted, which can prevent the blind search of particles and greatly improve the search efficiency. When performing alternating search in the present invention, the global optimal fitness function value is introduced for particle position initialization, making the particle distribution more reasonable and improving the search efficiency; a linearly varying constraint factor is introduced in the present invention to perform different constraints on angle search and distance search, preventing the search from being too divergent or concentrated due to the change in the order of magnitude of the search range; a linearly transformed inertia weight and self-learning factor are introduced in the present invention to improve the global search ability in the early stage and the local search ability in the later stage of the particle swarm optimization algorithm.

[0042] This method introduces the particle swarm optimization algorithm into the three-dimensional MUSIC sound source localization technology and improves the search strategy, making the calculation efficiency of the three-dimensional MUSIC sound source localization technology improved, and at the same time ensuring the localization accuracy under high signal-to-noise ratio.

[0043] Other advantages, objectives and features of the present invention will be described to some extent in the subsequent specification, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following specification. Brief Description of the Drawings

[0044] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be described in detail preferably with reference to the accompanying drawings, where:

[0045] Figure 1 is a schematic flow chart of the method of the present invention;

[0046] Figure 2 is a spatial spectrum slice diagram at different traversal positions; where (a) is r = 1.00 m; (b) is r = 1.50 m; (c) is r = 2.00 m; (d) is θ = 162.5°, (e) is θ = 163.0°, (f) is θ = 163.0°,

[0047] Figure 3 is the performance comparison result of the present invention with the MUSIC algorithm and the HOC-MUSIC algorithm; (a) is the root mean square error RMSE comparison of the positioning results under different signal-to-noise ratios; (b) is the cumulative distribution function CDF comparison of the misidentification rates of the positioning results under different signal-to-noise ratios; (c) is the positioning time comparison of different algorithms;

[0048] Figure 4Monte Carlo random test results for the performance of the present invention; (a) shows the root mean square error RMSE, the false recognition rate CDF, and the time used for positioning of the algorithm with the change in the number of sound sources; (b) shows the root mean square error RMSE, the false recognition rate CDF, and the time used for positioning of the algorithm with the change in the number of sampling beats; (c) shows the root mean square error RMSE, the false recognition rate CDF, and the time used for positioning of the algorithm with the change in the number of particles; (d) shows the root mean square error RMSE, the false recognition rate CDF, and the time used for positioning of the algorithm with the change in the number of iterations;

[0049] Figure 5 It is the diagram of the test positioning results; (a) is the single sound source positioning result; (b) is the double sound source positioning result; (c) is the triple sound source positioning result; (d) is the 3D diagram of the single sound source positioning result; (e) is the 3D diagram of the double sound source positioning result; (f) is the 3D diagram of the triple sound source positioning result. Detailed implementation manners

[0050] The following uses specific specific examples to illustrate the implementation manners of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. Without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0051] It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. Therefore, only the components related to the present invention are shown in the diagrams, rather than being drawn according to the number, shape, and size of the components in actual implementation. The type, quantity, and proportion of each component in its actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0052] In the following description, a large number of details are explored to provide a more thorough explanation of the embodiments of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present invention difficult to understand.

[0053] Embodiment 1:

[0054] Please refer to Figures 1 to 4, an embodiment of the present invention provides a three-dimensional MUSIC sound source localization method based on an improved particle swarm optimization algorithm. The simulation software used in this embodiment is MATLAB 2022b, and a microphone array with 128 array elements and a speaker of model Philips S1009 are used to transmit sound signals for verification. As Figure 1 shown, the method specifically includes the following steps:

[0055] Step S1: Construct a covariance matrix from the sound signal data collected by an array of M array elements and separate the noise subspace, which specifically includes the following steps:

[0056] S11: Construct a covariance matrix from the sound signal data collected by the array elements;

[0057] R = E[y(t)y(t) H

[0058] In the formula, R is the constructed covariance matrix, E represents taking the expectation, [·] H represents the transpose, and y(t) is the sound signal collected by the array;

[0059] S12: Perform eigenvalue decomposition on the covariance matrix. According to the number of sound sources, take all columns with a column number greater than the number of sound sources as the noise subspace;

[0060]

[0061] In the formula, U S represents the signal subspace, Σ S represents the eigenvalues corresponding to the signal subspace, U N represents the noise subspace, and Σ N represents the eigenvalues corresponding to the noise subspace.

[0062] Step S2: Construct the array manifold in three-dimensional space using larger step sizes for azimuth, elevation, and distance, which specifically includes the following steps: Construct the array manifold according to the near-field propagation model of the sound signal;

[0063] y(t) = A(f, Φ)s(t) + W(t)

[0064]

[0065] In the formula, A(f, Φ) = [a1(f, Φ1), a2(f, Φ2),..., a j (f, Φ N )], a j (f, Φ j ) represents the array manifold, f represents the sound signal frequency, θ, ​r represents the azimuth, elevation angle, and distance respectively, s(t) represents the sound source amplitude, W(t) represents the noise, and Γ ji represents the distance between the j-th sound source and the i-th microphone, and c represents the speed of sound.

[0066] Step S3: Perform preliminary localization of the sound source according to the MUSIC algorithm principle, which specifically includes the following steps: Use the orthogonality relationship between the array manifold and the noise subspace in the MUSIC algorithm principle for preliminary localization to obtain the preliminary localization result

[0067] P(f, Φ) = ||A(f, Φ) H U N || 2

[0068] where P(f, Φ) is the reciprocal of the spatial spectrum, and A(f, Φ) H is the constructed array manifold;

[0069] Step S4: Initialize the parameters of the particle swarm optimization algorithm for alternating search and search in the three-dimensional space, which specifically includes the following steps:

[0070] S41. Initialize the parameters of the particle swarm optimization algorithm: The maximum value K dmax and the minimum value K dmin of the linearly varying constraint factor, the maximum number of iterations iter max per single time, the maximum number of alternations change max , the maximum value ω max and the minimum value ω min of the linearly varying inertia weight, the changing self-learning factor c 1max , c 1middle , c 1min , the boundary ratios λ1 and λ2 that control the change of the self-learning sub-factor, and the condition γ for stopping iteration. Initialize the parameters to control the parameter changes of the particle swarm optimization algorithm during the iteration;

[0071] K d = K dmax -(K dmax - K dmin )·u / U

[0072] U = iter max ·change max

[0073] ω = ω max -(ω max - ω min )·u / U

[0074]

[0075] S42. Initialize the velocities of all particles;

[0076]

[0077] In the formula, is the velocity of the k-th particle at the u-th iteration in the d-th dimension, K d is the constraint factor, ω is the inertia weight, representing the influence of the velocity update of the particle in the previous iteration on the velocity update of the particle in the next iteration, rand3 is a random number between [0, 1], Value g is the overall optimal fitness value of the particle, which is P(f, Φ) at the initial positioning when searching for the first time;

[0078] S43. According to the formula, first traverse the angle θ, and then traverse the distance r;

[0079]

[0080] In the formula, c1 and c2 are the self-learning factor and social learning factor of the particle respectively, representing the influence of the particle's own historical best position and the historical best position of the entire particle swarm on the particle, rand1 and rand2 are random numbers between [0, 1], pBest k is the historical best fitness function value of the k-th particle itself, gBest is the historical best fitness function value of the entire particle swarm, is the position of the k-th particle at the u-th iteration in the d-th dimension;

[0081] S44. If gBest < γ or the maximum number of iterations is reached, then output the positioning result;

[0082] [[ID=�5]]S45. If there is no change in gBest in 10 consecutive traversals in one search, then swap the search dimension (angle / distance), return to S42, otherwise continue the traversal in the current dimension;

[0083] In this example, the beneficial effects of the present invention are demonstrated through simulation test verification and on-site test verification:

[0084] Algorithm comparison: As Figure 3As shown in (a)-(b), after SNR < -10 dB, the RMSE and CDF of the MUSIC algorithm and the PSO-MUSIC algorithm increase rapidly. When SNR = -30 dB, the RMSE reaches 0.470 m and 0.519 m respectively, and the CDF reaches 0.894 and 0.954 respectively. When SNR ≥ -10 dB, the RMSE and CDF are similar, the RMSE is less than 0.025 m, and the CDF is less than 0.025. Since the HOC-MUSIC algorithm can reduce the influence of white noise when constructing the high-order cumulant, the RMSE and CDF do not change much under the simulated SNR, and the average values are 0.103 m and 0.198 respectively. When SNR < -20 dB, compared with the MUSIC algorithm and the PSO-MUSIC algorithm, the HOC-MUSIC algorithm has a lower CDF but a higher RMSE. When SNR ≥ -20 dB, the CDF and RMSE are higher than those of the MUSIC algorithm and the PSO-MUSIC algorithm. From Figure 3 As shown in (c) of 1P , among the three algorithms, the PSO-MUSIC algorithm has the shortest positioning time, which is 3.1% and 28.9% of the MUSIC algorithm and the HOC-MUSIC algorithm respectively. Conduct a complexity analysis of the three algorithms. Let M be the number of microphones, T be the number of signal beats, k1, k2, k3 be the search grid numbers of the azimuth angle, elevation angle, and distance respectively, k 2P , k 3P are the search grid numbers corresponding to the preliminary positioning of the PSO-MUSIC algorithm respectively, n particles , iter max , change max are the number of particles, the maximum number of iterations, and the maximum number of alternations of the PSO-MUSIC algorithm respectively. The complexity of the MUSIC algorithm is mainly composed of covariance matrix construction, eigenvalue decomposition, and grid search, which is O(max(M 2 T, M 3 , k1k2k3M 2 )),. Compared with the MUSIC algorithm, the HOC-MUSIC algorithm decouples the angle search and the distance search, so its complexity is O(max(M 2 T, M 3 , (k1k2 + k3)M 2 )) The PSO-MUSIC algorithm conducts preliminary positioning and changes the grid search to a particle swarm search, and the complexity is O(max(M 2 T, M 3 , (k 1P k 2P k 3P + n particles iter max change max M2 ))). Generally, the number of general grids is much larger than the number of signal beats and the number of microphones. Therefore, the complexity of the MUSIC algorithm is O(k1k2k3M 2 ), the complexity of the HOC-MUSIC algorithm is O((k1k2 + k3)M 2 ), and the complexity of the PSO-MUSIC algorithm is O(k 1P k 2P k 3P +n particles iter max change max M 2 ). Since some search termination conditions are introduced in the PSO-MUSIC algorithm, the PSO-MUSIC algorithm generally does not reach the complexity upper limit in most cases.

[0085] Algorithm performance: Figure 4 In (a), it shows the performance of the PSO-MUSIC algorithm as the number of sound sources increases. When the number of located sound sources increases, the RMSE and CDF of the PSO-MUSIC algorithm gradually increase. When the number of sound sources is 6, the RMSE and CDF reach 0.120m and 0.196 respectively, and at this time the positioning effect of the algorithm is already poor. As the number of sound sources increases, the positioning time also increases proportionally. (b) shows the performance of the PSO-MUSIC algorithm as the number of beats increases. When the number of beats is equal to 640, the RMSE and CDF of the PSO-MUSIC algorithm are 0.015m and 0.023 respectively. After that, as the number of beats increases, the RMSE and CDF decrease slowly. And as the number of beats increases, within 1408 beats, the positioning time basically remains unchanged. (c) shows the performance of the PSO-MUSIC algorithm as the number of particles increases. As the number of particles increases, the RMSE and CDF gradually decrease, and the positioning time gradually increases. When the number of particles is 25, the RMSE and CDF are 0.016m and 0.024 respectively. When the number of particles is 40, the RMSE and CDF are 0.010m and 0.017 respectively, but the time consumed is 1.29 times that when the number of particles is 25. Considering the actual allowable positioning time, allowable positioning error, and positioning error rate, it is more appropriate to select the number of particles between 25 - 30. (d) shows the performance of the PSO-MUSIC algorithm as the maximum number of iterations increases. The PSO-MUSIC algorithm mentioned in this paper involves alternating search, and the number of iterations involves iter max and change max two parameters. During the simulation, make change max = 0.5·iter max , then It can be seen that when the maximum number of iterations reaches 648, the RMSE and CDF are 0.014 m and 0.021 respectively, and then the RMSE and CDF decrease slowly. At the same time, as the maximum number of iterations increases, the time consumed also gradually increases. It is more appropriate to select the maximum number of iterations between 450 and 648 by comprehensively considering the actual allowable time, the allowable positioning error, and the positioning error rate.

[0086] The present invention is respectively used to identify the sound source, and the results are as Figure 5 shown, where (a), (b), and (c) are the result diagrams of single sound source, double sound source, and triple sound source positioning respectively, and (d), (e), and (f) are the 3D diagrams of single sound source, double sound source, and triple sound source positioning. The x-y plane is the plane where the microphone array is located. The sampling beats of this experiment are set to 600 beats, and the filtering bandwidth is 200 Hz. The loudspeaker is randomly placed within the range of 0.2 m - 1.5 m from the center of the microphone array. The sound source frequencies played by the microphone are 14 kHz, 16 kHz, and 18 kHz from left to right in sequence. The RMSE of single sound source positioning is 0.018 m, the RMSE of double sound source positioning is 0.035 m, and the RMSE of triple sound source positioning is 0.035 m. It can be seen that when the PSO-MUSIC algorithm locates the three-dimensional position of the narrowband signal, it can accurately locate the sound source position. However, as the number of sound sources increases, the positioning RMSE will increase. The positioning times of single sound source, double sound source, and triple sound source on the mobile PC are 0.85 s, 1.36 s, and 2.33 s respectively. Therefore, the PSO-MUSIC algorithm mentioned in this paper can maintain high-precision three-dimensional positioning of narrowband signals with high efficiency.

[0087] From the results of the simulation test and the field test, it is proved that the sound source positioning ability of the present invention is better than that of the prior art.

[0088] Embodiment 2:

[0089] An electronic device, characterized in that it includes a memory and a processor;

[0090] The memory is used to store a computer program;

[0091] The processor is used to implement the three-dimensional MUSIC sound source positioning method based on the improved particle swarm optimization algorithm as described in Embodiment 1 when executing the computer program.

[0092] Embodiment 3:

[0093] A computer-readable storage medium, characterized in that a computer program is stored on the storage medium, and when the computer program is executed by a processor, the three-dimensional MUSIC sound source positioning method based on the improved particle swarm optimization algorithm as described in Embodiment 1 is implemented.

[0094] Embodiment 4:

[0095] A computer program product, characterized in that: it includes a computer program which, when executed by a processor, implements the three-dimensional MUSIC sound source localization method based on the improved particle swarm optimization algorithm as described in Embodiment 1.

[0096] In the above embodiments, the mention of "this embodiment" in the specification means that the specific features, structures or characteristics described in connection with the embodiments are included in at least some embodiments, but not necessarily all embodiments. Multiple occurrences of "this embodiment" do not necessarily all refer to the same embodiment.

[0097] In the above embodiments, although the present invention has been described in connection with specific embodiments of the present invention, many substitutions, modifications and variations of these embodiments will be apparent to those of ordinary skill in the art based on the foregoing description. For example, other storage structures (e.g., dynamic RAM (DRAM)) can be used with the embodiments discussed. Embodiments of the present invention are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims.

[0098] For the computer-readable storage medium in this embodiment, those of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to the computer program. The foregoing computer program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps including the above method embodiments; and the foregoing storage medium includes: various media such as ROM, RAM, magnetic disk or optical disk that can store program codes.

[0099] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication therebetween. The memory is used to store a computer program, the communication interface is used for communication, and the processor and the transceiver are used to run the computer program so that the electronic terminal executes each step of the above method.

[0100] In this embodiment, the memory may include a random access memory (Random Access Memory, abbreviated as RAM), and may also include a non-volatile memory, such as at least one disk memory.

[0101] The above-mentioned processor may be a general-purpose processor, including a central processing unit (CPU for short), a network processor (NP for short), etc.; it may also be a digital signal processor (DSP for short), an application specific integrated circuit (ASIC for short), a field-programmable gate array (FPGA for short), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0102] The present invention can be used in numerous general-purpose or special-purpose computing system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and so on.

[0103] The present invention can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present invention can also be practiced in a distributed computing environment, where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0104] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the present technical solution, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A three-dimensional MUSIC sound source localization method based on an improved particle swarm optimization algorithm, characterized in that: It includes the following steps: S1: Construct a covariance matrix from the acoustic signal data collected by an array of M elements and separate the noise subspace; S2: Construct an array manifold in three-dimensional space using large step sizes for azimuth, elevation, and distance; S3: Conduct a preliminary localization of the sound source according to the principle of the MUSIC algorithm; S4: Initialize the parameters of the particle swarm optimization algorithm for alternating search and search in three-dimensional space.

2. The three-dimensional MUSIC sound source localization method based on the improved particle swarm optimization algorithm according to claim 1, wherein: The step S1 includes the following steps: S11: Construct a covariance matrix from the acoustic signal data collected by the elements; R = E[y(t)y(t) H ​ where \(R\) is the constructed covariance matrix, \(E\) represents taking the expectation, and \([\cdot]^T\) H represents the transpose, and \(y(t)\) is the acoustic signal collected by the array; S12: Perform eigenvalue decomposition on the covariance matrix, and take all columns with a column number greater than the number of sound sources as the noise subspace according to the number of sound sources; where, U S represents the signal subspace, Σ S represents the eigenvalue corresponding to the signal subspace, U N represents the noise subspace, Σ N represents the eigenvalue corresponding to the noise subspace.

3. The three-dimensional MUSIC sound source localization method based on the improved particle swarm optimization algorithm according to claim 1, characterized in that: The step S2 includes the following steps: Construct an array manifold according to the near-field propagation model of the acoustic signal; y(t) = A(f, Φ)s(t) + W(t) where \(A(f,\varPhi)=[a_1(f,\varPhi_1),a_2(f,\varPhi_2),\cdots,a j (f,\varPhi N )]\), \(a j (f,\varPhi j )\) represents the array manifold, \(f\) represents the acoustic signal frequency; \(\theta\), \(r\) represent the azimuth angle, the elevation angle and the distance respectively; \(s(t)\) represents the sound source amplitude, \(W(t)\) represents the noise, \(\Gamma ji \) represents the distance between the \(j\)-th sound source and the \(i\)-th microphone, and \(c\) represents the speed of sound.

4. The three-dimensional MUSIC sound source localization method based on the improved particle swarm optimization algorithm according to claim 1, characterized in that: The S3 includes the following steps: Initial positioning is carried out by using the orthogonal relationship between the array manifold and the noise subspace in the MUSIC algorithm principle to obtain the initial positioning result P(f, Φ) = ||A(f, Φ) H U N || 2 where P(f, Φ) is the reciprocal of the spatial spectrum, and A(f, Φ) H is the constructed array manifold.

5. The three-dimensional MUSIC sound source localization method based on the improved particle swarm optimization algorithm according to claim 1, wherein: The S4 includes the following steps: S41: Initialize the parameters of the particle swarm optimization algorithm, including: the maximum value K of the constraint factor with linear variation dmax and the minimum value K dmin , the maximum number of iterations iter per single time max , the maximum number of alternations change max , the maximum value ω of the inertia weight with linear variation max and the minimum value ω min , the varying self-learning factor c 1max , c 1middle , c 1min , the boundary ratios λ1, λ2 for controlling the variation of the self-learning sub-factor, the condition γ for stopping iteration; Initialize the parameters to control the parameter variation of the particle swarm optimization algorithm during the iteration process: K d = K dmax -(K dmax - K dmin )·u / U U = iter max ·change max ω = ω max -(ω max -ω min )·u / U S42: Initialize the velocities of all particles; In the formula, is the velocity of the k-th particle at the u-th iteration in the d-th dimension, K d is the constraint factor, ω is the inertia weight, representing the influence of the velocity update of the particle in the previous iteration on the velocity update of the particle in the next iteration, rand3 is a random number between [0, 1], Value g is the overall optimal fitness value of the particle, which is P(f, Φ) at the initial positioning when the search is first carried out; S43: First, traverse the angle θ, and then traverse the distance r: Wherein, c1 and c2 are respectively the self-learning factor and the social learning factor of the particle, representing the influence of the historical best position of the particle itself and the historical best position of the entire particle swarm on the particle respectively, rand1 and rand2 are random numbers between [0, 1], and pBest k is the historical best fitness function value of the k-th particle itself, and gBest is the historical best fitness function value of the entire particle swarm. is the position of the k-th particle at the u-th iteration in the d-th dimension; S44: If gBest < γ or the maximum number of iterations is reached, output the localization result; S45: If there is no change in gBest after U consecutive traversals in one search, swap the search dimensions and return to S42, otherwise continue the traversal in the current dimension.

6. An electronic device, characterized in that, It includes a memory and a processor; The memory is used to store a computer program; The processor is used to implement the three-dimensional MUSIC sound source localization method based on the improved particle swarm optimization algorithm as described in any one of claims 1 to 5 when executing the computer program.

7. A computer-readable storage medium, characterized in that, A computer program is stored on the storage medium, and when the computer program is executed by the processor, it implements the three-dimensional MUSIC sound source localization method based on the improved particle swarm optimization algorithm as described in any one of claims 1 to 5.

8. A computer program product, characterized in that: It includes a computer program, and when the computer program is executed by the processor, it implements the three-dimensional MUSIC sound source localization method based on the improved particle swarm optimization algorithm as described in any one of claims 1 - 5.

Images