Multi-array sound source localization method based on particle swarm optimization algorithm

Abstract

The invention provides a multi-array sound source localization method based on a particle swarm optimization algorithm. The method comprises the following steps: calculating and obtaining estimated sound source coordinate parameters of a real sound source under each single microphone array; constructing a positioning direction line between the estimated sound source coordinate parameter of each single microphone array and the corresponding reference array element; constructing an optimal target function of the sum of the distances from the optimal sound source point to each positioning direction line after the positioning data of each single microphone array is fused; after the optimal target function of the single microphone array is solved and a sound source position calculation result of the single microphone array is obtained, performing iterative search on a plurality of sound source position parameters of the multi-array microphone array through a particle swarm optimization algorithm until an iteration stop condition is met; and outputting and taking the optimal position parameter as a positioning parameter of a real sound source. According to the invention, the problem of positioning failure of the single microphone array is effectively solved, and the positioning precision of the single microphone array is improved.

Description

Technical Field

[0001] The invention relates to the field of sound source localization, and in particular to a multi-array sound source localization method based on a particle swarm optimization algorithm. Background Art

[0002] Sound source localization is a method based on acoustic principles that uses electronic devices to obtain sound wave information during sound propagation to determine the location information of the target sound source. In recent years, with the development of digital signal processing technology, sound source localization has been widely used in the fields of civilian robot positioning, industrial equipment noise detection, military weapon equipment positioning, and cochlear implants. It plays an important role in positioning, especially in cases of blurred vision and electromagnetic wave shielding, and can achieve all-round positioning in space.

[0003] The key issue of sound source localization lies in the research of microphone array structure and positioning algorithm. By optimizing the microphone array structure and the performance of the positioning algorithm, the real-time performance and accuracy of the sound source localization system can be improved. The microphone array is composed of two or more microphones arranged in a certain geometric structure, which is used to extract the sound wave information during the propagation of the sound source. For the sound source localization estimation algorithm, the most widely used one is the sound source localization algorithm based on the delay difference. This algorithm constructs a set of equations related to the sound source position according to the delay difference between the sound signals, and obtains the spatial three-dimensional position information of the sound source by solving the set of equations.

[0004] Existing sound source localization technologies mostly use single microphone arrays, and their positioning algorithms are only applicable to single microphone arrays. Affected by factors such as the spacing between microphones, the three-dimensional coordinate parameters of the sound source (distance, azimuth, pitch angle), and noise signals, the sound source position parameter errors actually calculated by the single array positioning method are large, making it difficult to achieve high-precision positioning. Increasing the number of microphone arrays can improve positioning accuracy, but the method requires an increased amount of calculation, which ultimately leads to a decrease in the real-time performance of the system, making it difficult to solve the problem that a single microphone array cannot accurately locate the sound source. Summary of the invention

[0005] In order to solve the problems existing in the background technology, the present invention proposes a multi-array sound source localization method based on a particle swarm optimization algorithm.

[0006] A multi-array sound source localization method based on particle swarm optimization algorithm comprises the following steps:

[0007] S100, establishing a multi-array microphone array model and a unified coordinate system, selecting one of the single microphone arrays as a system reference array, and selecting a reference array element of each single microphone array;

[0008] S200, calculating and obtaining the estimated sound source coordinate parameters of the real sound source in each single microphone array, converting the estimated sound source coordinate parameters in each single microphone array into the estimated sound source coordinate parameters in the system reference array, that is, setting the position of the reference array element in the reference array as the origin of the multi-array system coordinate system, and converting the estimated sound source coordinate parameters in each single array small system coordinate system into the corresponding sound source estimated coordinate parameters in the reference array coordinate system through coordinate transformation;

[0009] S300, constructing a positioning direction line between the estimated sound source coordinate parameters of each single microphone array and the corresponding reference array element;

[0010] S400, constructing an optimal objective function of the sum of the distances from the optimal sound source point to each positioning direction line after the fusion of the positioning data of each single microphone array;

[0011] S500, after solving the optimal objective function of the single microphone array and obtaining the sound source position calculation result of the single microphone array, iteratively search the multiple sound source position parameters of the multi-array microphone array through the particle swarm optimization algorithm until the iteration stop condition is reached, and output and use the optimal position parameters as the positioning parameters of the real sound source.

[0012] Based on the above, in step S100, the multi-array microphone array includes multiple single microphone arrays, each single microphone array includes multiple microphones, and the reference array element is a microphone in the single microphone array; wherein the microphone is a shared microphone, and the shared microphone is cross-multiplexed to construct a single microphone array.

[0013] Based on the above, in step S200, according to the principle of sound source localization based on time delay difference, the Cartesian coordinate parameters and spherical coordinate system parameters of the estimated sound source of the real sound source under each single microphone array are calculated, and the Cartesian coordinate parameters are converted into parameters under the system reference array.

[0014] Based on the above, the estimated Cartesian coordinate position parameters of the real sound source S under the first single microphone array A1 are S1(x1, y1, z1), and the spherical coordinate position parameters are S1(r1, θ1, φ1). According to the principle of sound source localization based on arrival time difference, the following set of equations is constructed:

[0015]

[0016] The conversion relationship between the Cartesian coordinate system and the spherical coordinate system in the first single-microphone array is:

[0017]

[0018] The coordinates of the real sound source S in the system reference array are S(x, y, z), and the coordinates in the spherical coordinate system are The distance between the real sound source S and the first single-microphone array reference element M0 is r1; the sound wave propagation speed of the real sound source S is c; the single-microphone array A j The acoustic path difference of the signal received by each array element relative to the array reference array element in (j=1, 2, 3, 4, 5, 6, 7, 8) is d ji =c*τ ji , (i=1, 2, 3, 4), τ ji represents the signal delay difference between the i-th microphone in the j-th array and its corresponding reference microphone in the single array, j represents the array number of the single-microphone array, i represents the array element number in each single-microphone array; the spacing between each microphone array element is D.

[0019] Based on the above, in step S300, the position parameters of the real sound source are S (x, y, z), and the estimated sound source coordinate parameters of each single microphone array are S j (x j ,y j , z j ), the positioning direction line constructed by each single microphone array is the positioning direction vector in A is a single microphone array j The reference array element.

[0020] Based on the above, in step S400, the positioning coordinate parameters of the optimal sound source point after the fusion of the positioning data of each single microphone array are S0 (x0, y0, z0), and the distance between S0 and the positioning direction line of each single microphone array is d j , calculated by the following formula:

[0021]

[0022] in, represents the direction vector of the positioning direction line, Indicates S0 and S j The vector between them; then the optimal objective function with the minimum distance sum as the indicator is expressed by the following function:

[0023]

[0024] Among them, f j (x0, y0, z0) = d j The variables are (x0, y0, z0).

[0025] Based on the above, in step S500, the speed and position of each particle in each dimension are updated according to equations 12 and 13:

[0026]

[0027] p k (t+1)=p k (t)+v k (t+1) (13)

[0028] Among them, t is the number of iterations, and the particle position parameter p k =[x k ,y k , z k ] is the position parameter of the sound source, rand1 and rand2 are two different random numbers, the number of particles is k = 300, the maximum number of iterations is N = 100, the variable dimension is d = 3, the inertia coefficient ω = 0.5, and the random factor c1 = c2 = 0.5; the optimal position parameter of a single particle represents the optimal position of the kth particle; the global optimal position parameter gbes t This is the sound source localization result.

[0029] The present invention has outstanding substantive features and significant progress compared with the prior art. Specifically, the present invention applies a data fusion method to sound source localization, uses the rough sound source position parameters calculated by each single microphone array, and constructs an optimal objective function between the optimal target sound source position parameters and the rough sound source position parameters in combination with geometric relationships, and determines the solution range of the sound source position parameters to be optimized according to the actual positioning rough results, and then uses a particle swarm optimization algorithm to iteratively solve the objective function, outputs the last iteration global optimal position parameters of the particle swarm optimization algorithm, which are the precise position parameters of the target sound source, and realizes high-precision positioning of the sound source, effectively overcomes the positioning failure problem of the single microphone array, and when the number of microphone array elements is the same, efficiently utilizes the sound source signal collected by each array element, and performs sound source localization, and then fuses the sound source localization results of multiple single microphone arrays, which improves the positioning accuracy of the single microphone array itself, improves the problem that the sound source localization result is greatly affected by the failed single array in the traditional existing fusion method, and improves the positioning accuracy of the single microphone array. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic diagram of the structure of the microphone array model of the present invention.

[0031] Figure 2 It is a schematic block diagram of the process of the present invention. DETAILED DESCRIPTION

[0032] 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.

[0033] like Figure 2 As shown in FIG. 1 , a multi-array sound source localization method based on a particle swarm optimization algorithm is described as follows:

[0034] A multi-array microphone array model and a unified coordinate system are established. The multi-array microphone array includes multiple single microphone arrays, and each single microphone array includes multiple microphones. Among them, each microphone in the single microphone array is an array element in the single microphone array; the microphone is shared, and the single microphone array is constructed by cross-multiplexing, that is, a microphone can be an array element in different single microphone arrays at the same time. Select one of the single microphone arrays as the system reference array, and select the reference array element of each single microphone array. The reference array element of the system reference array is the system reference array element. Specifically:

[0035] like Figure 1 As shown, in this embodiment, the multi-array microphone array model includes 14 microphones, and M0, M1, M2, M3, M4, M5, M6, M7, M8, M9, M10, M11, M12 and M13 represent the 14 microphones respectively, and an XYZ Cartesian three-dimensional coordinate system is established. Assuming that the spacing between each microphone array element is D, the coordinates of each microphone can be expressed as M0(0, 0, 0), M1(D, 0, 0), M2(0, D, 0), M3(0, 0, D), M4(2D, 0, 0), M5(2D, D, 0), M6(2D, 0, D), M7(2D, 2D, 0), M8(D, 2D, 0), M9(2D, 2D, D), M10(0, 2D, 0), M11(0, 2D, D), M12(D, D, 0), M13(D, D, D).

[0036] In this embodiment, the first four-element single-microphone array A1 is constructed with M0, M1, M2, and M3, and the reference array element of the single-microphone array A1 is M0. The second four-element single-microphone array A2 is constructed with M4, M5, M1, and M6, and the reference array element of the single-microphone array A2 is M4. The third four-element single-microphone array A3 is constructed with M7, M8, M9, and M5, and its reference array element is M7. The fourth four-element single-microphone array A4 is constructed with M10, M2, M8, and M11, and its reference array element is M10. The fifth four-element single-microphone array A5 is constructed with M12, M8, M2, and M13, and its reference array element is M12. The sixth four-element single-microphone array A6 is constructed with M12, M2, M1, and M13. The seventh four-element single-microphone array A7 is constructed with M12, M1, M5, and M13. The eighth four-element single-microphone array A8 is constructed with M12, M5, M8, and M13. The fifth, sixth, seventh, and eighth single-microphone arrays A5, A6, A7, and A8 use M12 as the reference array element, respectively. The first array A1 is selected as the system reference array, and the system reference array element is M0.

[0037] Assume that the coordinates of the real sound source S in the system reference array are S(x, y, z), and the coordinates in the spherical coordinate system are The distance between the sound source and the system reference array element M0 is r, the projection of S on the XOY plane is A′, the elevation angle AOZ is θ, the azimuth angle A′OX is φ, the real sound source emits sound waves in the form of spherical waves, and the propagation speed is c. It can be obtained that the microphone array A j The acoustic path difference of the signal received by each array element in (j=1, 2, 3, 4, 5, 6, 7, 8) relative to its array reference array element is d ji =c*τ ii , (i=1, 2, 3, 4), j represents the ordinal number of the single microphone array, and i represents the ordinal number of the array element in each array.

[0038] According to the principle of sound source localization based on time delay difference, the sound source localization calculation formula of the single microphone array is derived, and the estimated sound source Cartesian coordinate parameters and spherical coordinate system parameters of the real sound source under each single microphone array are calculated. The estimated sound source Cartesian coordinate parameters are converted into parameters under the reference array, which are specifically:

[0039] Taking the first single microphone as an example, the Cartesian coordinate parameters and spherical coordinate parameters of the single microphone array estimated sound source S1 are derived.

[0040] Assuming that the Cartesian coordinate position parameters of the sound source S estimated under the first single microphone array A1 are S1(x1, y1, z1), and the spherical coordinate position parameters are S1(r1, θ1, φ1), the following set of equations can be constructed according to the principle of sound source localization based on arrival time difference:

[0041]

[0042] The conversion relationship between the Cartesian coordinate system and the spherical coordinate system in the first array A1 is:

[0043]

[0044] Simplifying the equation group (1) yields the following relationship group:

[0045]

[0046] Substituting the relational expression group (2) into the first equation of the equation group (1), we can obtain the following quadratic equation:

[0047]

[0048] Solving the quadratic equation can obtain the distance r1 between the sound source estimated by the first array A1 and the coordinate origin. Substituting the value r1 into the relationship group (2) can obtain the Cartesian coordinate parameters S1 (x1, y1, z1) of the estimated sound source estimated by the first array A1.

[0049] Combining equations (2) and (3), we can obtain the relationship between the azimuth parameter φ1 and the elevation parameter θ1 of the estimated sound source in the first array A1:

[0050]

[0051] Substituting the distance r1, we can obtain the spherical coordinate parameters S1(r1, θ1, φ1) of the sound source localization of the first array A1.

[0052] Thus, the first single-microphone array A1 is obtained to perform sound source positioning estimation on the real sound source S(x, y, z), and the estimated sound source Cartesian coordinate parameters are S1(x1, y1, z1), and the spherical coordinate parameters are S1(r1, θ1, φ1).

[0053] Next, we transform S into j (x j ,y j , z j ) is converted into the Cartesian position parameter S′ under the system reference array A1 j (x′ j , y′ j , z′ j ).Depend on Figure 1It can be seen that the position relationship structure of the single microphone array A1, A2, A3, A4, A5, A6, A7, and A8 is obtained by rotating a certain angle α with the Z axis as the axis and then translating according to the vector λ. z for:

[0054]

[0055] The coordinate transformation relationship is:

[0056] S′ j =S j *R+λ (7)

[0057] From the above, we can know that the single microphone array A1 is the system reference array, that is, the A1 array is rotated by α1=0° relative to the reference array according to the right-hand relationship, and then translated according to the vector λ=M0-M0=(0,0,0). After the coordinate transformation, the position parameters of the sound source localization position of the first microphone array in the reference array coordinate system are:

[0058] S′1=S1*R1+λ1 (8)

[0059] in:

[0060]

[0061] Analogously to the sound source estimation calculation process of the first microphone array A1, the sound path difference d of the first microphone array A1 is 11 , d 12 , d 13 , d 14 Replaced by the sound path difference d corresponding to the second single microphone array A2 21 , d 22 , d 23 , d 24 , the Cartesian coordinate parameters S2(x2, y2, z2) and spherical coordinate parameters S2(r2, θ2, φ2) of the estimated sound source S2 of the second single microphone array A2 are derived. According to the right-hand system relationship, the array conversion angle is α2=90°, and the translation vector is λ2=M4-M0=(D, 0, 0). Substituting these parameters into formula (6), the positioning parameters S′2(x′2, y′2, z′2) of the second microphone array A2 in the reference array Cartesian coordinate system can be obtained.

[0062] Analogously to the sound source estimation calculation process of the first microphone array A1, the sound path difference d of the first microphone array A1 is 11 , d 12 , d 13 , d 14Replaced by the sound path difference d corresponding to the third single microphone array A3 31 , d 32 , d 33 , d 34 , the Cartesian coordinate parameters S3 (x3, y3, z3) and spherical coordinate parameters S3 (r3, θ3, φ3) of the estimated sound source S3 of the third single microphone array A3 are derived. According to the right-hand system relationship, the array conversion angle is α3 = 180°, and the translation vector is λ3 = M7-M0 = (2D, 2D, 0). Substituting these parameters into formula (6), the positioning parameters S′3 (x′3, y′3, z′3) of the third microphone array A3 in the reference array Cartesian coordinate system can be obtained.

[0063] Analogously to the sound source estimation calculation process of the first microphone array A1, the sound path difference d of the first microphone array A1 is 11 , d 12 , d 13 , d 14 Replaced by the sound path difference d corresponding to the fourth single microphone array A4 41 , d 42 , d 43 , d 44 , the Cartesian coordinate parameters S4 (x4, y4, z4) and spherical coordinate parameters S4 (r4, θ4, φ4) of the estimated sound source S4 of the fourth single microphone array A4 are derived. According to the right-hand system relationship, the array conversion angle is α4 = 270°, and the translation vector is λ4 = M7-M0 = (0, 2D, 0). Substituting these parameters into formula (6), the positioning parameters S′4 (x′4, y′4, z′4) of the fourth microphone array A4 in the reference array Cartesian coordinate system can be obtained.

[0064] Analogously to the sound source estimation calculation process of the first microphone array A1, the sound path difference d of the first microphone array A1 is 11 , d 12 , d 13 , d 14 Replaced by the sound path difference d corresponding to the fifth single microphone array A5 51 , d 52 , d 53 , d 54 , the Cartesian coordinate parameters S5(x5, y5, z5) and spherical coordinate parameters S5(r5, θ5, φ5) of the estimated sound source S5 of the fifth single microphone array A5 are derived. According to the right-hand system relationship, the array conversion angle is α5=90°, and the translation vector is λ5=M12-M0=(D, D, 0). Substituting these parameters into formula (6), the positioning parameters S′5(x′5, y′5, z′5) of the fifth microphone array A5 in the reference array Cartesian coordinate system can be obtained.

[0065] Analogously to the sound source estimation calculation process of the first microphone array A1, the sound path difference d of the first microphone array A1 is 11 , d 12 , d 13 , d 14 Replaced by the sound path difference d corresponding to the sixth single microphone array A6 61 , d 62 , d 63 , d 46 , the Cartesian coordinate parameters S6 (x6, y6, z6) and spherical coordinate parameters S6 (r6, θ6, φ6) of the estimated sound source S6 of the sixth single microphone array A6 are derived. According to the right-hand system relationship, the array conversion angle is α6 = 180°, and the translation vector is λ6 = M12-M0 = (D, D, 0). Substituting these parameters into formula (6), the positioning parameters S′6 (x′6, y′6, z′6) of the sixth microphone array A6 in the reference array Cartesian coordinate system can be obtained.

[0066] Analogously to the sound source estimation calculation process of the first microphone array A1, the sound path difference d of the first microphone array A1 is 11 , d 12 , d 13 , d 14 Replaced by the sound path difference d corresponding to the seventh single microphone array A7 71 , d 72 , d 73 , d 47 , the Cartesian coordinate parameters S7(x7, y7, z7) and spherical coordinate parameters S7(r7, θ7, φ7) of the estimated sound source S7 of the seventh single microphone array A7 are derived. According to the right-hand system relationship, the array conversion angle is α7=270°, and the translation vector is λ7=M12-M0=(D, D, 0). Substituting these parameters into formula (6), the positioning parameters S′7(x′7, y′7, z′7) of the seventh microphone array A7 in the reference array Cartesian coordinate system can be obtained.

[0067] Analogously to the sound source estimation calculation process of the first microphone array A1, the sound path difference d of the first microphone array A1 is 11 , d 12 , d 13 , d 14 Replaced by the sound path difference d corresponding to the eighth single microphone array A8 81 , d 82 , d 83 , d 48, the Cartesian coordinate parameters S8(x8, y8, z8) and spherical coordinate parameters S8(r8, θ8, φ8) of the estimated sound source S8 of the eighth single microphone array A8 are derived. According to the right-hand system relationship, the array conversion angle is α8=360°, and the translation vector is λ8=M12-M0=(D, D, 0). Substituting these parameters into formula (6), the positioning parameters S8(x8, y8, z8) of the eighth microphone array A8 in the reference array Cartesian coordinate system are obtained.

[0068] Construct the direction line between each single microphone array sound source coordinate parameter and the corresponding array reference element, which is specifically:

[0069] The estimated sound source position parameters obtained by using the eight single microphone arrays obtained above to estimate the sound source location for the real sound source with the real position parameters S(x, y, z) are S j (x j ,y j , z j ), using the position parameter to construct the positioning direction vector of the array in A is a single microphone array j The reference array element.

[0070] Construct the optimal objective function of the sum of the distances from the optimal sound source point to each direction line after fusion, which is specifically:

[0071] Assume that the coordinate parameters of the sound source positioning after the fusion of the single microphone array positioning data are S0 (x0, y0, z0). When the single microphone array positioning result is the same as the true value of the sound source position (there is no positioning error), the eight positioning direction lines can intersect at one point, which is the true position of the sound source. At this time, the sum of the distances from point S0 to the eight positioning direction lines is zero. Due to the errors in the positioning system, the eight positioning direction lines cannot intersect at one point, so the optimal objective function with the minimum distance sum as the indicator can be constructed.

[0072] Assume that the distance between S0 and the positioning direction line of each single microphone array is d j , can be calculated by the following formula:

[0073]

[0074] in, Indicates S0 and S j The vector between them, then the optimal objective function with the minimum distance sum as the indicator can be expressed by the following function:

[0075]

[0076] Among them, f j (x0, y0, z0) = d jThe variables are (x0, y0, z0).

[0077] According to the sound source position calculation results of the single microphone array, the parameters of the particle swarm optimization algorithm are initialized. The particle swarm optimization algorithm is used to iteratively search for the precise sound source position parameters until the iteration stop condition is reached and the optimal position parameters are output, which are specifically:

[0078] Step 1: According to S j (x j ,y j , z j ), find the average location parameter

[0079] Step 2: Initialize the particle swarm position boundaries as follows:

[0080] 2.1) The upper and lower bounds of the x-coordinate position are

[0081] 2.2) The upper and lower bounds of the coordinate y position are

[0082] 2.3) The upper and lower bounds of the coordinate z position are

[0083] 2.4) The upper and lower bounds of the search speed in the x direction are (v xu , v xl )=0.1*(x u , x l ).

[0084] 2.4) The upper and lower bounds of the search speed in the y direction are (v yu , v yl )=0.1*(y u ,y l ).

[0085] 2.4) The upper and lower bounds of the search speed in the z direction are (v zu , v zl )=0.1*(z u , z l ).

[0086] Step 3: Randomly initialize the particle position p within the range limit k =[x k ,y k , z k ],speed is the initial solution of the particle swarm optimization problem, the number of particles is k = 300, the maximum number of iterations is N = 100, the variable dimension is d = 3, the inertia coefficient ω = 0.5, and the random factor c1 = c2 = 0.5.

[0087] Step 4: Calculate the fitness function value F of all particles. According to the requirements of the objective function, compare the fitness value F of the previous generation and the next generation of a single particle, and update the optimal position parameter of a single particle. g best It is the optimal position parameter set of all particles in the current iteration.

[0088] Step 5: According to the objective function requirements, when the optimal fitness value corresponding to the optimal position parameters of all particles is better than the global optimal fitness value, update the global optimal position parameter g best And the global optimal fitness value F.

[0089] Step 6: Update the velocity and position of each particle in each dimension according to Equation 12 and Equation 13 respectively.

[0090]

[0091] p k (t+1)=p k (t)+v k (t+1) (13)

[0092] Among them, t is the number of iterations, and the particle position parameter p k =[x k ,y k , z k ] is the position parameter of the sound source, rand1 and rand2 are two different random numbers.

[0093] Step 7: Repeat steps 4-6 until the maximum number of iterations or the set fitness value is reached, exit the loop iteration, and output the global optimal position parameters.

[0094] The fused Cartesian coordinates are converted into spherical coordinates according to equations 1 and 5.

[0095] The global optimal position parameter finally output in the particle swarm optimization algorithm is the final sound source localization result of the multi-array sound source localization data fusion method based on the particle swarm optimization algorithm proposed in this embodiment. The technical solution of the present invention effectively overcomes the problem of positioning failure of a single microphone array. When the number of microphone array elements is the same, the sound source signal collected by each array element is efficiently utilized to perform sound source localization, and then the sound source localization results of multiple single microphone arrays are fused. This improves the positioning accuracy of the single microphone array, improves the problem that the sound source localization results are greatly affected by the failed single array in the traditional existing fusion method, and improves the positioning accuracy of the single microphone array.

[0096] In this embodiment, the initial sound source is simulated by using the multi-array sound source localization data fusion method based on the particle swarm optimization algorithm and the original eight single four-element array algorithms. The specific simulation results of the three times are shown in Tables 1-3, where the corresponding polar coordinate parameters are (distance r, azimuth angle θ, pitch angle φ):

[0097] Table 1 shows the simulation results of sound source localization for the real polar coordinates of the sound source (6m, 45°, 47°):

[0098] Distance r / m Pitch angle θ / ° Azimuth φ / ° Single Microphone Array A1 5.87 45.00 47.06 Single Microphone Array A2 5.83 45.47 46.74 Single Microphone Array A3 6.02 45.00 46.93 Single Microphone Array A4 5.83 44.53 46.74 Single Microphone Array A5 6.05 45.00 47.03 Single Microphone Array A6 5.83 45.19 46.72 Single Microphone Array A7 5.90 45.00 46.82 Single Microphone Array A8 5.83 44.81 46.72 Average positioning results 5.89 45.00 46.85 Fusion positioning results 5.97 45.00 46.97

[0099] Table 1 Table 2 shows the simulation results of sound source localization for the real polar coordinates of the sound source (8.5m, 55°, 15°):

[0100] Distance r / m Pitch angle θ / ° Azimuth φ / ° Single Microphone Array A1 8.36 54.94 15.23 Single Microphone Array A2 9.28 54.35 15.02 Single Microphone Array A3 8.81 55.00 15.10 Single Microphone Array A4 7.83 54.22 15.09 Single Microphone Array A5 8.18 54.91 15.02 Single Microphone Array A6 9.43 54.85 14.87 Single Microphone Array A7 8.50 55.13 14.98 Single Microphone Array A8 8.27 54.94 15.01 Average positioning results 8.58 54.79 15.04 Fusion positioning results 8.52 55.00 15.02

[0101] Table 2 Table 3 shows the simulation results of sound source localization for the real polar coordinates of the sound source (4m, 22°, 55°):

[0102] Distance r / m Pitch angle θ / ° Azimuth φ / ° Single Microphone Array A1 3.98 22.17 54.80 Single Microphone Array A2 4.04 21.28 55.00 Single Microphone Array A3 3.98 22.01 54.79 Single Microphone Array A4 4.13 22.10 55.94 Single Microphone Array A5 4.78 20.61 56.60 Single Microphone Array A6 4.43 18.79 56.41 Single Microphone Array A7 3.88 23.13 54.03 Single Microphone Array A8 3.89 21.74 54.48 Average positioning results 4.14 21.47 55.26 Fusion positioning results 3.98 21.99 54.80

[0103] Table 3

[0104] As shown in Table 1-3, the calculation accuracy of the sound source spherical coordinates of the multi-array sound source localization data fusion method based on the particle swarm optimization algorithm in this embodiment is higher than that of the single microphone array, and the accuracy is higher than the average positioning result of eight single microphone arrays. The final simulation error of the multi-array sound source localization data fusion method based on the particle swarm optimization algorithm is low. The simulation results show that the multi-array sound source localization data fusion method based on the particle swarm optimization algorithm has a high sound source localization accuracy, which is 0-97.8% higher than the single microphone array distance localization accuracy, 0-98.7% higher than the elevation angle localization accuracy, and 0-99.8% higher than the average positioning result of eight single microphone arrays. The distance localization accuracy is improved by 50.1%, the elevation angle localization accuracy is improved, and the azimuth angle localization accuracy is improved by 23.1%-80.3%. Therefore, the multi-array sound source localization data fusion method based on the particle swarm optimization algorithm has a better sound source localization effect.

[0105] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A multi-array sound source localization method based on particle swarm optimization algorithm, characterized in that: Includes steps: S100, establishing a multi-array microphone array model and a unified coordinate system, selecting one of the single microphone arrays as a system reference array, and selecting a reference array element of each single microphone array; S200, calculating and obtaining estimated sound source coordinate parameters of the real sound source under each single microphone array, and converting the estimated sound source coordinate parameters under each single microphone array into estimated sound source coordinate parameters under the system reference array; S300, constructing a positioning direction line between the estimated sound source coordinate parameters of each single microphone array and the corresponding reference array element; S400, constructing an optimal objective function of the sum of the distances from the optimal sound source point to each positioning direction line after the fusion of the positioning data of each single microphone array; S500, after solving the optimal objective function of the single microphone array and obtaining the sound source position calculation result of the single microphone array, iteratively search the multiple sound source position parameters of the multi-array microphone array through the particle swarm optimization algorithm until the iteration stop condition is reached, and output and use the optimal position parameters as the positioning parameters of the real sound source.

2. The multi-array sound source localization method based on particle swarm optimization algorithm according to claim 1, characterized in that: In step S100, the multi-array microphone array includes multiple single microphone arrays, each single microphone array includes multiple microphones, and the reference array element is a microphone in the single microphone array; wherein the microphone is a shared microphone, and the shared microphone is cross-multiplexed to construct a single microphone array.

3. The multi-array sound source localization method based on particle swarm optimization algorithm according to claim 1, characterized in that: In step S200, according to the principle of sound source localization based on time delay difference, the Cartesian coordinate parameters and spherical coordinate system parameters of the estimated sound source of the real sound source under each single microphone array are calculated, and the Cartesian coordinate parameters are converted into parameters under the system reference array.

4. The multi-array sound source localization method based on particle swarm optimization algorithm according to claim 3 is characterized in that: The estimated Cartesian coordinate position parameters of the real sound source S under the first single microphone array A1 are S1(x1, y1, z1), and the spherical coordinate position parameters are S1(r1, θ1, φ1). According to the principle of sound source localization based on arrival time difference, the following set of equations is constructed: The conversion relationship between the Cartesian coordinate system and the spherical coordinate system in the first single-microphone array is: The coordinates of the real sound source S in the system reference array are S(x, y, z), and the coordinates in the spherical coordinate system are The distance between the real sound source S and the first single-microphone array reference element M0 is r1; the sound wave propagation speed of the real sound source S is c; the single-microphone array A j The acoustic path difference of the signal received by each array element relative to the array reference element in (j=1,2,3,4,5,6,7,8) is d ji =c*τ ji , (i=1,2,3,4), τ ji represents the signal delay difference between the i-th microphone in the j-th array and its corresponding reference microphone in the single array, j represents the array number of the single-microphone array, i represents the array element number in each single-microphone array; the spacing between each microphone array element is D.

5. The multi-array sound source localization method based on particle swarm optimization algorithm according to claim 1, characterized in that: In step S300, the position parameters of the real sound source are S (x, y, z), and the estimated sound source coordinate parameters of each single microphone array are S j (x j ,y j ,z j ), the positioning direction line constructed by each single microphone array is the positioning direction vector in A is a single microphone array j The reference array element.

6. The multi-array sound source localization method based on particle swarm optimization algorithm according to claim 1, characterized in that: In step S400, the positioning coordinate parameters of the optimal sound source point after the fusion of the positioning data of each single microphone array are S0 (x0, y0, z0), and the distance between S0 and the positioning direction line of each single microphone array is d j , calculated by the following formula: in, represents the direction vector of the positioning direction line, Indicates S0 and S j The vector between them; then the optimal objective function with the minimum distance sum as the indicator is expressed by the following function: Among them, f j (x0,y0,z0)=d j The variables are (x0,y0,z0).

7. The multi-array sound source localization method based on particle swarm optimization algorithm according to claim 1, characterized in that: In step S500, the velocity and position of each particle in each dimension are updated according to equations 12 and 13: p k (t+1)=p k (t)+v k (t+1) (13) Among them, t is the number of iterations, and the particle position parameter p k =[x k ,y k ,z k ] is the position parameter of the sound source, rand1 and rand2 are two different random numbers, the number of particles is k = 300, the maximum number of iterations is N = 100, the variable dimension is d = 3, the inertia coefficient ω = 0.5, and the random factor c1 = c2 = 0.5; the optimal position parameter of a single particle represents the optimal position of the kth particle; the global optimal position parameter g best This is the sound source localization result.

Images