Sound source localization method based on improved particle swarm optimization

Abstract

The invention provides a sound source localization method based on an improved particle swarm algorithm. Comprising the following steps: selecting a sound source positioning scene, acquiring a measurement data set by using a microphone array, and establishing a mathematical model; initializing the positions and speeds of the particles, and calculating the fitness value of each particle to evaluate the distance between the current position of the particle and the target solution; setting an inertia weight to control the influence of the current speed of the particles on the new speed, and setting an individual learning factor and a social learning factor to control and coordinate the behaviors of the particles in the search space; setting an iteration control condition, combining with a hybrid optimization algorithm, and utilizing a boundary processing strategy to carry out iteration updating of particles; and after iteration updating is completed, error analysis is performed on a sound source positioning result, and the particle swarm optimization algorithm is continuously optimized by using an optimization strategy according to an analysis result. The method can effectively improve the sound source positioning precision and algorithm performance, is suitable for a complex acoustic environment, and provides an efficient and feasible solution for the application in the acoustic field.

Description

Technical Field

[0001] The invention relates to the technical field of sound source localization, and in particular to a sound source localization method based on an improved particle swarm algorithm. Background Art

[0002] The sound source localization method is a technology that determines the location of the sound source by analyzing the sound signal. Existing sound source localization methods mainly include: 1) Based on the arrival time difference method, the sound source position is calculated by the arrival time difference of the sound signal recorded by multiple microphone arrays, but it has the defect of being sensitive to noise and reverberation; 2) Based on the proportional method, the incident angle of the sound source is calculated by the microphone array, but it requires more microphone arrays and has high computational complexity; 3) Based on the energy method, the energy difference measured by the sound source at different microphones is used for positioning, but it is sensitive to environmental changes and the positioning accuracy is limited by distance and obstacles; 4) Based on machine learning and neural network methods, it requires a large amount of training data and the model has poor interpretability.

[0003] Particle Swarm Optimization (PSO) is an optimization algorithm based on swarm intelligence, proposed by Kennedy and Eberhart in 1995. It simulates the foraging behavior of bird flocks and searches for the optimal solution through information sharing between individuals. Each "particle" represents a potential solution, and the particle updates its position and velocity by following the current best particle to move in the search space. It is relatively easy to implement sound source localization using the particle swarm algorithm, has strong adaptability to sound source localization problems, has low difficulty in solving problems in complex environments, and exhibits good global search capabilities in multi-peak and complex acoustic environments, making it suitable for solving nonlinear and multi-modal sound source localization problems.

[0004] However, directly applying the existing PSO algorithm to sound source localization will have the following problems: 1) The global search capability is limited and it is easy to fall into the local optimal trap. 2) The algorithm parameters lack an adaptive adjustment mechanism, resulting in a low convergence speed. 3) The ability to adapt to dynamic environments is poor, and the response is slow or loses effectiveness. 4) The diversity of the particle population is lost, affecting the exploration of the global optimal solution. 5) Improper boundary constraint processing, when the search space is limited and the particles exceed the allowable range, the lack of a good processing mechanism will cause the algorithm to be invalid. Therefore, the present invention proposes a sound source localization method based on an improved particle swarm algorithm to solve the above problems. Summary of the invention

[0005] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a sound source localization method based on an improved particle swarm algorithm, which can effectively improve the sound source localization accuracy and algorithm performance, adapt to complex acoustic environments, and provide an efficient and feasible solution for applications in the acoustic field.

[0006] To achieve the above object, the present invention provides the following solution: a sound source localization method based on an improved particle swarm algorithm, comprising the following steps:

[0007] Select a sound source localization scenario, collect sound source signals using a microphone array and perform data preprocessing on the sound source signals to obtain a measurement data set, and establish a mathematical model based on the microphone array and the measurement data set;

[0008] Initialize the position and velocity of the particle, and calculate the fitness value of each particle to evaluate the distance between the particle's current position and the target solution;

[0009] Set the inertia weight to control the influence of the particle's current velocity on the new velocity, and set the individual learning factor and social learning factor to control and coordinate the behavior of particles in the search space;

[0010] Iterative control conditions are set and combined with a hybrid optimization algorithm to iteratively update particles, and a boundary processing strategy is used to keep particles running in the search space during the iterative update process;

[0011] After completing the iterative update, the error analysis of the sound source localization results is performed, and the particle swarm algorithm is continuously optimized using the optimization strategy based on the analysis results.

[0012] Optionally, using a microphone array to collect a sound source signal and performing data preprocessing on the sound source signal to obtain a measurement data set, and establishing a mathematical model based on the microphone array and the measurement data set, including:

[0013] Recording the timestamp of each microphone receiving the signal and calculating the time difference of the received signal based on the timestamp to obtain an initial data set, and performing denoising, filtering and formatting on the initial data set to obtain a measurement data set;

[0014] Selecting a base point as a reference origin in the selected sound source localization scene, establishing a three-dimensional coordinate system based on the reference origin, and arranging and recording the position coordinates of each microphone in the three-dimensional coordinate system to complete the establishment of a mathematical model;

[0015] Based on the measurement data set, the time difference between the microphone receiving signals is calculated to construct an error function for evaluating the quality of the particle position, and the directional arrival angle of the measurement signal is used to evaluate the prediction error of the sound source direction.

[0016] Optionally, initialize the particle's position and velocity, including:

[0017] Search all position ranges of the sound source based on the three-dimensional coordinate system to determine a three-dimensional search space;

[0018] Generating initial positions of a plurality of particles randomly and uniformly in the three-dimensional search space and obtaining initial position coordinates {x, y, z}, and then selecting a set number of particles for searching based on the generated plurality of particles;

[0019] Based on the three-dimensional search space and the selected set number of particles, the initial speed of each particle and the maximum speed and minimum speed of the particle in each dimension are set, and the initial speed is dynamically adjusted according to the size of the search space and the required number of iterations.

[0020] Optionally, calculate the fitness value of each particle to evaluate the distance between the particle's current position and the target solution, including:

[0021] Obtaining and integrating the current position of each particle in the three-dimensional search space and the actual data in the measurement data set;

[0022] Predict and calculate the sound source localization index of each particle's current position based on the integrated data; the sound source localization index includes a time difference or a direction angle;

[0023] The sound source localization index and actual data are compared and calculated to obtain a prediction error, and the prediction error is then used to define a fitness function of the particle to obtain a fitness value; wherein the prediction error is inversely proportional to the fitness value, and the smaller the prediction error, the larger the fitness value.

[0024] Optionally, set the inertia weight to control the influence of the particle's current velocity on the new velocity, and set the individual learning factor and social learning factor to control and coordinate the behavior of particles in the search space, including:

[0025] Defining the type and range of inertia weight, and selecting a linear decrease strategy so that the inertia weight decreases linearly with the number of iterations; the inertia weight includes a high inertia weight for search and a low inertia weight for development;

[0026] In each iteration, the current inertia weight is calculated according to the current iteration number, and the current velocity of the particle is updated according to the current inertia weight;

[0027] When updating the current velocity of a particle, the individual learning factor and the social learning factor are used to make the particle find a balance between its own best historical position and the group's best global position.

[0028] Optionally, the iterative update process of particles includes:

[0029] Using the current inertia weight, the individual learning factor and the social learning factor to update the speed of each particle, and then using the updated particle speed and boundary processing strategy to update the particle position;

[0030] Calculate the current fitness value of each particle's new position, and compare the particle's current fitness value with the historical best fitness value. If the current fitness value is better than the historical best fitness value, update the particle's historical best position to obtain a new individual optimal position.

[0031] Combined with a hybrid optimization algorithm, the new individual optimal positions of all particles are compared and the optimal particle is selected to update the global optimal position; the hybrid optimization algorithm includes a genetic algorithm and simulated annealing for enhancing the global search capability and diversity of the algorithm;

[0032] In each iterative update process, the iteration information is recorded and the change of the fitness value is monitored; the iteration information includes the global best fitness value, the average fitness value and the particle position distribution.

[0033] Optionally, the iteration control conditions include:

[0034] Before the iterative update begins, an initial condition and a stop condition are set; the initial condition includes a maximum number of iterations, a minimum error threshold, and a convergence standard, and the stop condition includes a maximum number of iterations stop, a minimum error threshold stop, and a no-change convergence stop;

[0035] A control flow is set during the iterative update process; the control flow includes initializing an iteration counter, starting iteration, checking a stop condition, iteration counting, logging, and terminating output.

[0036] Optionally, a boundary processing strategy is used to keep particles running in the search space during the iterative update process, including: defining the boundary of the three-dimensional search space, and checking the particle position after each iterative update. If there are particles beyond the boundary, the particles are bounced or reset back to the boundary by using a particle bounce boundary method, a particle position reset method, a periodic boundary condition method, or a boundary random reset method.

[0037] Optionally, after completing the iterative update, an error analysis is performed on the sound source localization result, including:

[0038] Define the error metric in sound source localization, then collect the actual position and estimated position of each test position to perform error calculation and obtain error data;

[0039] Analyzing the error data using statistical tools to obtain the distribution, skewness and kurtosis of the error data, and making an error histogram based on the analysis results to observe the error concentration trend and dispersion;

[0040] The accuracy of sound source localization is evaluated based on the error data and the analysis results thereof, and a comparison diagram of the predicted position and the actual position is generated to intuitively display the positioning deviation.

[0041] Optionally, the optimization strategy includes fuzzy logic for dynamically adjusting inertia weights and learning factors, an environmental simulation model for adjusting algorithm adaptability, and multi-particle swarm parallelization for rapid exploration;

[0042] The multi-particle swarm parallelization creates multiple particle swarms, allowing the algorithm to explore in multiple search areas and share information between different particle swarms to find the global optimal position.

[0043] The present invention discloses the following technical effects by providing a sound source localization method based on an improved particle swarm algorithm:

[0044] 1. The present invention can effectively build the basic environment required by the particle swarm algorithm by collecting data and establishing a mathematical model, and the meticulous processing of the coordinate system, objective function and data set ensures that the subsequent sound source localization process is more accurate and efficient.

[0045] 2. The present invention provides a good start for the subsequent particle swarm optimization iterative process by appropriately initializing the position and speed of the particles. At the same time, by calculating the fitness value of each particle, the distance between the current position of the particle and the target solution is evaluated, which helps to identify the relative advantages and disadvantages of the particles in the global search space, and guides the adjustment of the particle speed and position during the iteration process, thereby gradually converging to the true sound source position.

[0046] 3. By setting important parameters such as inertia weight and learning factor, the present invention can control the impact of the current speed on the new speed, adjust the global exploration ability and local development ability of the algorithm, and determine the degree of dependence of a particle on its own historical best position and the global best position of the group when updating the speed, thereby improving the algorithm's ability to balance exploration and utilization in the search space, effectively controlling and coordinating the behavior of particles in the search space, and enabling the algorithm to more efficiently locate the sound source.

[0047] 4. Iterative update is the core content of the algorithm, which can continuously optimize the state of the particle group, gradually approach the actual sound source position, and realize efficient and accurate sound source positioning. At the same time, the present invention introduces hybrid optimization algorithms, boundary processing strategies and a variety of advanced optimization strategies and technologies during the iterative optimization process and after the optimization is completed, which can achieve the following effects: improving the global search ability of the algorithm, avoiding falling into local optimality, dynamically adjusting parameters, improving environmental adaptability, avoiding particles crossing the boundary, etc., so that the algorithm's performance in improving the accuracy and speed of sound source positioning is significantly improved.

[0048] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0050] Figure 1 A schematic diagram of a method flow chart provided by an embodiment of the present invention;

[0051] Figure 2 A schematic diagram of a particle iterative update process provided by an embodiment of the present invention;

[0052] Figure 3 A schematic diagram of a flow chart of particle iteration control conditions provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0055] like Figure 1-3 As shown, the present invention provides a sound source localization method based on an improved particle swarm algorithm, comprising the following steps:

[0056] 1. Select a sound source localization scenario, use a microphone array to collect sound source signals and perform data preprocessing on the sound source signals to obtain a measurement data set, and establish a mathematical model based on the microphone array and the measurement data set; the specific process is:

[0057] 1.1 Record the timestamp of each microphone receiving the signal and calculate the time difference of the received signal based on the timestamp to obtain an initial data set, and perform denoising, filtering and formatting on the initial data set to obtain a measurement data set;

[0058] 1.2 Select a base point in the selected sound source localization scene as the reference origin defined as (0,0,0), and establish a three-dimensional coordinate system based on the reference origin; usually choose a right-hand coordinate system, the X axis points to a fixed direction (such as east), the Y axis points to another orthogonal direction (such as north), and the vertical upward direction is the Z axis; set the coordinate system unit and scale, usually in meters or centimeters, and adjust the appropriate scale range according to the size of the site;

[0059] 1.3 Arrange and record the position coordinates of each microphone in the three-dimensional coordinate system to complete the establishment of the mathematical model;

[0060] 1.4 Based on the measurement data set, the time difference between the microphone receiving signals is calculated to construct an error function for evaluating the quality of the particle position, and the directional arrival angle of the measurement signal is used to evaluate the prediction error of the sound source direction.

[0061] Through the above process, the basic environment required by the particle swarm algorithm can be effectively constructed. This meticulous processing of the coordinate system, objective function and data set ensures that the subsequent sound source localization process is more accurate and efficient.

[0062] 2. Initialize the position and velocity of the particle, and calculate the fitness value of each particle to evaluate the distance between the current position of the particle and the target solution.

[0063] 2.1 Initialize the position and velocity of the particle, including:

[0064] 2.11 Search all position ranges of the sound source based on the three-dimensional coordinate system to determine a three-dimensional search space (i.e., a search space on the x, y, and z axes);

[0065] 2.12 Randomly and uniformly generate the initial positions of multiple particles in the three-dimensional search space (the initial positions should be evenly distributed in the entire search space to ensure global search capability) and obtain the initial position coordinates {x, y, z}, and then select a set number of particles from the generated multiple particles for search. The more particles there are, the better the coverage of the exploration space, but the computational cost increases. It is usually appropriate to select 20 to 50 particles;

[0066] 2.13 Based on the three-dimensional search space and the selected set number of particles, set the initial speed of each particle and the maximum and minimum speeds of the particles in each dimension, and dynamically adjust the initial speed according to the size of the search space and the required number of iterations.

[0067] By executing the above steps, the position and velocity of the particles are properly initialized, providing a good start for the subsequent particle swarm optimization iterative process.

[0068] 2.2 Calculate the fitness value of each particle, including:

[0069] 2.21 obtaining the current position of each particle in the three-dimensional search space and the actual data in the measurement data set (including the signal time or direction angle received by the microphone) and integrating them;

[0070] 2.22 predicting and calculating the sound source localization index of each particle's current position based on the integrated data; the sound source localization index includes a time difference or a direction angle;

[0071] 2.23 Compare and calculate the sound source localization index and the actual data to obtain a prediction error E (the sum of squares of the errors is usually used to evaluate the fitness), and then use the prediction error E to define the fitness function of the particle to obtain a fitness value F; wherein the prediction error is inversely proportional to the fitness value (F=1 / (1+E)), the smaller the prediction error, the larger the fitness value, and the particle exploration tends to the optimal solution.

[0072] Through the above steps, the current fitness value can be calculated for each particle. This calculation method helps to identify the relative advantages and disadvantages of particles in the global search space, and guides the adjustment of particle speed and position during the iteration process, so as to gradually converge to the true sound source position.

[0073] 3. Set the inertia weight to control the impact of the particle's current velocity on the new velocity, and set the individual learning factor and social learning factor to control and coordinate the behavior of the particle in the search space; the specific process is:

[0074] 3.1 Define the type and range of inertia weight. The inertia weight is usually set within a range, such as from 0.9 (high inertia, heavy exploration) to 0.4 (low inertia, heavy development). A larger initial value allows the particle swarm to cover a larger search space in the early stage and find the optimal solution from the global perspective. A smaller final value allows the particles to refine the search for a more local optimal solution in the later stage;

[0075] 3.2 And select a suitable linear decreasing strategy to make the inertia weight decrease linearly with the number of iterations. Or according to experience or problem requirements, you can choose exponential decreasing, cosine function decreasing or other forms of nonlinear functions to adjust the change speed of the inertia weight;

[0076] 3.3 In each iteration, the current inertia weight is calculated according to the current number of iterations, and the current velocity of the particle is updated according to the current inertia weight;

[0077] 3.4 When updating the current velocity of particles, individual learning factors and social learning factors are used to make particles find a balance between their own best historical position and the group's best global position. Among them:

[0078] The individual learning factor is a factor that motivates each particle to move toward its own historical best position. It mainly controls the self-exploration ability of particles. A larger individual learning factor will encourage particles to pay more attention to their own discoveries, and an excessively large individual learning factor may cause particles to fall into a local optimum.

[0079] The social learning factor is a factor that motivates each particle to move toward the global optimal position. It mainly controls the degree of dependence of particles on group experience. A larger social learning factor makes particles more inclined to move toward the global optimal solution, which is conducive to rapid convergence, but too large a social learning factor may cause premature convergence.

[0080] Therefore, in general, the individual learning factor and the social learning factor are set in the range of (0,4), and can be dynamically adjusted during the algorithm execution process according to the problem requirements.

[0081] Through the design of the above parameters, the influence of the current particle speed on the new speed can be controlled, thereby adjusting the global exploration ability and local development ability of the algorithm; at the same time, the behavior of particles in the search space can be effectively controlled and coordinated, so that the algorithm can locate the sound source more efficiently and improve the performance and stability of the overall algorithm.

[0082] 4. If Figure 2-3 As shown, the iterative control conditions are set and combined with the hybrid optimization algorithm to perform iterative update of particles, and the boundary processing strategy is used to keep the particles running in the search space during the iterative update process.

[0083] 4.1 As Figure 2 As shown in Figure 2, the iterative update process of particles includes:

[0084] 4.11 using the current inertia weight, the individual learning factor and the social learning factor to update the speed of each particle, and then using the updated particle speed and boundary processing strategy to update the particle position;

[0085] 4.12 Calculate the current fitness value of each particle's new position (usually calculated based on the proximity of the particle's current position to the measured sound source), and compare the particle's current fitness value with the historical best fitness value. If the current fitness value is better than the historical best fitness value, update the particle's historical best position to obtain a new individual optimal position.

[0086] 4.13 Combined with a hybrid optimization algorithm, compare the new individual optimal positions of all particles and select the best particle to update the global optimal position; the hybrid optimization algorithm includes a genetic algorithm and simulated annealing for enhancing the global search capability and diversity of the algorithm;

[0087] 4.14 Record iteration information and monitor changes in fitness values ​​during each iteration update; the iteration information includes the global optimal fitness value, the average fitness value, and the particle position distribution.

[0088] 4.2 As Figure 3 As shown, the iteration control conditions include:

[0089] 4.21 Before the iterative update begins, the initial conditions and stop conditions are set; the initial conditions include the maximum number of iterations, the minimum error threshold and the convergence standard, and the stop conditions include the maximum number of iterations, the minimum error threshold and the no-change convergence stop;

[0090] 4.22 Set up a control flow during the iterative update process; the control flow includes initializing an iteration counter, starting iteration, checking stop conditions, iteration counting, logging, and terminating output.

[0091] 4.3 Boundary processing strategies include:

[0092] The boundary of the three-dimensional search space is defined, and the particle position after each iteration update is checked. If there are particles beyond the boundary, the particles are bounced or reset back to the boundary by using a particle bounce boundary method, a particle position reset method, a periodic boundary condition method, or a boundary random reset method.

[0093] 5. After completing the iterative update, perform error analysis on the results of sound source localization, and use optimization strategies to continuously optimize the particle swarm algorithm based on the analysis results.

[0094] 5.1 Perform error analysis, including:

[0095] 5.11 Error metrics are defined in sound source localization. Commonly used error metrics include: position error (the Euclidean distance between the estimated sound source position and the actual position), mean absolute error (the average of all estimated position errors), mean square error (the average of the squares of all position errors), and root mean square error (the square root of MSE, which provides the standard deviation of the error).

[0096] 5.12 Collect the actual position and estimated position of each test position to perform error calculation and obtain error data; then use statistical tools to analyze the error data to obtain the distribution, skewness and kurtosis of the error data, and make an error histogram based on the analysis results to observe the error concentration trend and dispersion;

[0097] 5.13 Based on the error data and the analysis results thereof, the accuracy of the sound source localization is evaluated, and a comparison diagram of the predicted position and the actual position is generated to intuitively display the positioning deviation.

[0098] 5.2 Optimization Strategy

[0099] Fuzzy logic: Dynamically adjust inertia weights and learning factors;

[0100] Environmental simulation model: The introduction of environmental model simulation allows the algorithm to adapt to changing environmental conditions. For example, adaptive adjustment of dynamic obstacles and sound reflection scenes;

[0101] Multi-particle swarm parallelization: By creating multiple particle swarms and running different initialized populations in parallel, the algorithm can explore multiple search areas together, and allow information sharing between different particle swarms, promote complementary advantages and features, and find the global optimal position more quickly.

[0102] The optimization strategy of the present invention includes but is not limited to the above method to improve the accuracy and speed of sound source localization, improve the performance of the algorithm, and improve the applicability and adaptability of the algorithm to complex acoustic environments.

[0103] Therefore, the present invention provides a sound source localization method based on an improved particle swarm algorithm, which can effectively improve the sound source localization accuracy and algorithm performance, adapt to complex acoustic environments, and provide an efficient and feasible solution for applications in the acoustic field.

[0104] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0105] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A sound source localization method based on an improved particle swarm algorithm, characterized in that: The following steps are involved: Select a sound source localization scenario, collect sound source signals using a microphone array and perform data preprocessing on the sound source signals to obtain a measurement data set, and establish a mathematical model based on the microphone array and the measurement data set; Initialize the position and velocity of the particle, and calculate the fitness value of each particle to evaluate the distance between the particle's current position and the target solution; Set the inertia weight to control the influence of the particle's current velocity on the new velocity, and set the individual learning factor and social learning factor to control and coordinate the behavior of particles in the search space; Iterative control conditions are set and combined with a hybrid optimization algorithm to iteratively update particles, and a boundary processing strategy is used to keep particles running in the search space during the iterative update process; After completing the iterative update, the error analysis of the sound source localization results is performed, and the particle swarm algorithm is continuously optimized using the optimization strategy based on the analysis results.

2. The sound source localization method based on the improved particle swarm algorithm according to claim 1, characterized in that: A microphone array is used to collect a sound source signal and data preprocessing is performed on the sound source signal to obtain a measurement data set, and a mathematical model is established based on the microphone array and the measurement data set, including: Recording the timestamp of each microphone receiving the signal and calculating the time difference of the received signal based on the timestamp to obtain an initial data set, and performing denoising, filtering and formatting on the initial data set to obtain a measurement data set; Selecting a base point as a reference origin in the selected sound source localization scene, establishing a three-dimensional coordinate system based on the reference origin, and arranging and recording the position coordinates of each microphone in the three-dimensional coordinate system to complete the establishment of a mathematical model; Based on the measurement data set, the time difference between the microphone receiving signals is calculated to construct an error function for evaluating the quality of the particle position, and the directional arrival angle of the measurement signal is used to evaluate the prediction error of the sound source direction.

3. The sound source localization method based on the improved particle swarm algorithm according to claim 2, characterized in that: Initialize the particle's position and velocity, including: Search all position ranges of the sound source based on the three-dimensional coordinate system to determine a three-dimensional search space; Generating initial positions of a plurality of particles randomly and uniformly in the three-dimensional search space and obtaining initial position coordinates {x, y, z}, and then selecting a set number of particles for searching based on the generated plurality of particles; Based on the three-dimensional search space and the selected set number of particles, the initial speed of each particle and the maximum speed and minimum speed of the particle in each dimension are set, and the initial speed is dynamically adjusted according to the size of the search space and the required number of iterations.

4. The sound source localization method based on the improved particle swarm algorithm according to claim 3, characterized in that: The fitness value of each particle is calculated to evaluate the distance between the particle's current position and the target solution, including: Obtaining and integrating the current position of each particle in the three-dimensional search space and the actual data in the measurement data set; Predict and calculate the sound source localization index of each particle's current position based on the integrated data; the sound source localization index includes a time difference or a direction angle; The sound source localization index and actual data are compared and calculated to obtain a prediction error, and the prediction error is then used to define a fitness function of the particle to obtain a fitness value; wherein the prediction error is inversely proportional to the fitness value, and the smaller the prediction error, the larger the fitness value.

5. The sound source localization method based on improved particle swarm algorithm according to claim 4, characterized in that: Set the inertia weight to control the influence of the particle's current velocity on the new velocity, and set the individual learning factor and social learning factor to control and coordinate the behavior of particles in the search space, including: Defining the type and range of inertia weight, and selecting a linear decrease strategy so that the inertia weight decreases linearly with the number of iterations; the inertia weight includes a high inertia weight for search and a low inertia weight for development; In each iteration, the current inertia weight is calculated according to the current iteration number, and the current velocity of the particle is updated according to the current inertia weight; When updating the current velocity of a particle, the individual learning factor and the social learning factor are used to make the particle find a balance between its own best historical position and the group's best global position.

6. The sound source localization method based on improved particle swarm algorithm according to claim 5, characterized in that: The iterative update process of particles includes: Using the current inertia weight, the individual learning factor and the social learning factor to update the speed of each particle, and then using the updated particle speed and boundary processing strategy to update the particle position; Calculate the current fitness value of each particle's new position, and compare the particle's current fitness value with the historical best fitness value. If the current fitness value is better than the historical best fitness value, update the particle's historical best position to obtain a new individual optimal position. Combined with a hybrid optimization algorithm, the new individual optimal positions of all particles are compared and the optimal particle is selected to update the global optimal position; the hybrid optimization algorithm includes a genetic algorithm and simulated annealing for enhancing the global search capability and diversity of the algorithm; In each iterative update process, the iteration information is recorded and the change of the fitness value is monitored; the iteration information includes the global best fitness value, the average fitness value and the particle position distribution.

7. The sound source localization method based on improved particle swarm algorithm according to claim 6, characterized in that: Iteration control conditions include: Before the iterative update begins, an initial condition and a stop condition are set; the initial condition includes a maximum number of iterations, a minimum error threshold, and a convergence standard, and the stop condition includes a maximum number of iterations stop, a minimum error threshold stop, and a no-change convergence stop; A control flow is set during the iterative update process; the control flow includes initializing an iteration counter, starting iteration, checking a stop condition, iteration counting, logging, and terminating output.

8. The sound source localization method based on improved particle swarm algorithm according to claim 7, characterized in that: The boundary processing strategy is used to keep the particles running in the search space during the iterative update process, including: defining the boundary of the three-dimensional search space, and checking the particle position after each iterative update. If there are particles beyond the boundary, the particles are bounced or reset back to the boundary by using a particle bounce boundary method, a particle position reset method, a periodic boundary condition method or a boundary random reset method.

9. The sound source localization method based on improved particle swarm algorithm according to claim 8, characterized in that: After the iterative update is completed, the error analysis of the sound source localization results is performed, including: Define the error metric in sound source localization, then collect the actual position and estimated position of each test position to perform error calculation and obtain error data; Analyzing the error data using statistical tools to obtain the distribution, skewness and kurtosis of the error data, and making an error histogram based on the analysis results to observe the error concentration trend and dispersion; The accuracy of sound source localization is evaluated based on the error data and the analysis results thereof, and a comparison diagram of the predicted position and the actual position is generated to intuitively display the positioning deviation.

10. The sound source localization method based on improved particle swarm algorithm according to claim 9, characterized in that: The optimization strategy includes fuzzy logic for dynamically adjusting inertia weights and learning factors, an environmental simulation model for adjusting algorithm adaptability, and multi-particle swarm parallelization for rapid exploration; The multi-particle swarm parallelization creates multiple particle swarms, allowing the algorithm to explore in multiple search areas and share information between different particle swarms to find the global optimal position.

Images