Multi-objective optimization method and system for loudspeaker array with sound field partition control

Abstract

The application relates to a loudspeaker array multi-objective optimization method and system for sound field partition control. According to a listening expectation, a sound field area is divided into a bright area and a dark area. A candidate point of loudspeaker arrangement is taken as a decision variable, and a decision variable vector is constructed. A multi-objective optimization model about the decision variable vector is constructed. A multi-objective genetic algorithm is adopted to solve the multi-objective optimization model to obtain a Pareto optimal solution. Based on an expected value of the multi-objective, a target function value of the multi-objective corresponding to each Pareto optimal solution, and weight coefficients of negative deviation variables and positive deviation variables of the target function value relative to the expected value, a linear weighted target programming model is constructed. The linear weighted target programming model is solved to obtain a Pareto optimal solution corresponding to a target function value when the linear weighted target programming model takes a minimum value, and optimal positions and quantities of the loudspeaker array are obtained. The application realizes the best balance among multiple targets and further improves the sound field partition effect.

Description

Technical Field

[0001] The present application relates to the technical field of sound field control, and in particular to a multi-objective optimization method and system for a loudspeaker array for sound field zoning control. Background Art

[0002] In recent years, with the advancement of science and technology and the improvement of living standards, automotive audio and video systems have attracted increasing attention. Passengers are increasingly concerned about the function of independent sound zones in the cabin's acoustic environment. Passengers in different seats expect to have different acoustic environments without interfering with each other. To achieve this personalized acoustic experience, it is necessary to solve the problem of in-vehicle speaker distribution. This problem, also known as the in-vehicle zone sound field control problem, is to recreate the target sound field in a specific desired area (the bright area) while reducing the impact of the target sound field on other areas (the dark area).

[0003] Speaker array optimization aims to design the optimal speaker distribution scheme, taking into account vehicle manufacturing costs, vehicle physical model, and user experience. To mitigate manufacturing costs, automakers consistently seek to provide personalized, independent sound zones using the fewest possible speakers. Currently, most speaker array optimization methods utilize the aforementioned metrics to formulate a single objective function to achieve optimal design, leaving room for improvement in the effectiveness of independent sound zones. Summary of the Invention

[0004] The embodiments of the present application provide a multi-objective optimization method and system for a loudspeaker array with sound field zoning control, which achieves an optimal balance between multiple objectives and further improves the sound field zoning effect.

[0005] In a first aspect, a multi-objective optimization method for a loudspeaker array with sound field partition control is provided, comprising:

[0006] According to listening expectations, the sound field area is divided into bright area and dark area;

[0007] Taking the candidate speaker placement points as decision variables, the decision variable vector is constructed according to the 0-1 type planning;

[0008] Constructing a multi-objective optimization model for the decision variable vector, the multi-objectives including: a first objective based on the contrast of acoustic potential energy in the bright area and the dark area, a second objective based on the number of loudspeakers, and a third objective based on the normalized mean square error in the bright area and the spatial consistency of the loudspeaker signals in the bright area;

[0009] Using a multi-objective genetic algorithm to solve the multi-objective optimization model to obtain several Pareto optimal solutions;

[0010] Constructing a linear weighted goal programming model based on the expected values ​​of the multiple objectives, the objective function values ​​of the multiple objectives corresponding to each of the Pareto optimal solutions, and the weight coefficients of the negative deviation variables and the positive deviation variables of the objective function values ​​relative to the expected values;

[0011] The linear weighted objective programming model is solved to obtain a Pareto optimal solution corresponding to the objective function value when the linear weighted objective programming model takes a minimum value, and the optimal position and number of the loudspeaker array are obtained.

[0012] In some embodiments, the decision variable vector is denoted as X, which includes:

[0013] X=[x1,x2,…,x j ,…,x K ]

[0014] Among them, x j =0 or 1, when x j = 0, no speaker is set at the jth speaker placement candidate point. j =1, the jth speaker placement candidate point is used to place a speaker, j=1, 2, ..., K, where K is the total number of speaker placement candidate points in the sound field area.

[0015] In some embodiments, the sound field area is divided into four cabin areas symmetrical about the center of the roof according to the position of the cabin in the car, so that x j′ =x j′+L =x j′+2L =x j′+3L ;

[0016] Wherein, j′=1, 2, …, L, L is the total number of candidate points for speaker placement in each cabin area, and K=4L.

[0017] In some embodiments, the multi-objective optimization model is as follows:

[0018]

[0019] Among them, F1(X) is the first target based on the contrast of sound potential energy in the bright area and the dark area, F2(X) is the second target based on the number of loudspeakers, and F3(X) is the third target based on the normalized mean square error in the bright area and the spatial consistency of the loudspeaker signal in the bright area.

[0020] In some embodiments, the contrast of the acoustic potential energy of the bright area and the dark area is recorded as AC i (X), and AC i (X) is:

[0021]

[0022] Among them, G B(f i ) is the transfer function matrix from the loudspeaker to the bright area microphone, with dimension M B ×K,M B is the number of microphones in the bright area, G B H (f i ) is G B (f i )'s conjugate transposed matrix;

[0023] G D (f i ) is the transfer function matrix from the loudspeaker to the dark zone microphone, with a dimension of M D ×K,M D is the number of dark area microphones, G D H (f i ) is G D (f i )’s conjugate transposed matrix; M B +M D =M, where M is the total number of microphones in the sound field;

[0024] q i is the driving signal vector of the loudspeaker, with a dimension of K×1, where K is the total number of candidate points for loudspeaker placement within the sound field area, and q i H q i The conjugate transposed matrix of ;

[0025] q i =[x1q1(f i ),x2q2(f i ),…,x j q j (f i ),…,x K q K (f i )] T

[0026] Where j = 1, 2, ..., K, q j (f i ) is the driving signal of the jth speaker, i is the frequency point order, f i is the frequency of the i-th frequency point, i=1,2,…, f min is the lower limit of the speaker frequency, f max is the upper limit of the speaker frequency, f i+1 -f i =1Hz, f i =f min ,f min +1,fmin +2,…,f max ;

[0027] Based on the contrast of acoustic potential energy in bright and dark areas AC i The first objective F1(X) is:

[0028]

[0029] In some embodiments, the second target F2(X) based on the number of speakers is:

[0030] F2(X)=||X||1

[0031] Here, ||·||1 is the 1-norm.

[0032] In some embodiments, the normalized mean square error of the bright area is recorded as E r,i (X), and E r,i (X) is:

[0033]

[0034] Among them, P BT (f i ) is the target sound field matrix for bright area reconstruction, is the square of the 2-norm;

[0035] The spatial consistency of the loudspeaker signal in the bright area is recorded as ΔL i (X), and ΔL i (X) is:

[0036] ΔL i (X) = L i,B,max (X)-L i,B,min (X)

[0037]

[0038] Among them, L i,B (X) is the sound pressure level of the microphone in the bright area, P ref (f i ) is the reference sound pressure, L i,B,max (X) is the L of all microphones in the bright area i,B The maximum value of (X), L i,B,min (X) is the L of all microphones in the bright area i,B (X) minimum value;

[0039] The third objective F3(X) based on the normalized mean square error in the bright area and the spatial consistency of the loudspeaker signal in the bright area is:

[0040]

[0041] In some embodiments, the method further includes determining a transfer function matrix, and determining the transfer function matrix specifically includes:

[0042] Place speakers at all candidate speaker placement points, and place microphones in bright and dark areas;

[0043] giving a test drive signal to each of the speakers one by one, and obtaining a measured sound pressure collected by each microphone;

[0044] Based on the test drive signal and the measured sound pressure, the transfer function matrix G(f i ):

[0045]

[0046] Among them, p 0，mj (f i ) is the sound pressure measured by the m-th microphone at the j-th speaker placement candidate point, m = 1, 2, ..., M, q 0，j (f i ) is the test driving signal of the jth speaker, is the transfer function from the jth loudspeaker to the mth microphone;

[0047] According to the microphones in the bright area and the dark area, the transfer function matrix G(f i ) to obtain the transfer function matrix G from the loudspeaker to the bright area microphone B (f i ) and the transfer function matrix G from the loudspeaker to the dark area microphone D (f i ).

[0048] In some embodiments, the method further includes optimizing the driving signal, and optimizing the driving signal specifically includes:

[0049] Construct the drive signal optimization model as follows:

[0050]

[0051] Solve the driving signal optimization model to obtain the driving signal vector q corresponding to the speaker with the maximum eigenvalue i * .

[0052] In some embodiments, optimizing the driving signal further includes:

[0053] q i * After multiplying by the coefficient c, the optimization model for the coefficient c is constructed as follows:

[0054]

[0055] in, is the square of the 2-norm, P BT (f i ) is the target sound field matrix for bright area reconstruction;

[0056] Solve the optimization model about coefficient c to obtain the optimal solution c of coefficient c * , change q i * Multiply by c * After that, we get c * q i * , and as the driving signal vector q of the loudspeaker i .

[0057] In some embodiments, the linear weighted goal programming model is:

[0058]

[0059] Among them, F a (X)+s a -t a =b a ,a=1,2,3,X is the decision variable vector. When a=1, F1(X) is the first target based on the contrast of the sound potential energy in the bright area and the dark area. When a=2, F2(X) is the second target based on the number of loudspeakers. When a=3, F3(X) is the third target based on the normalized mean square error in the bright area and the spatial consistency of the loudspeaker signal in the bright area. a ,t a ≥0,s a ×t a =0,b a F a The expected value of (X), s a F a (X) positive deviation variable, u a is a positive deviation variable s a The weight of t a F a (X) negative deviation variable, v a is a negative deviation variable t a The weight, w a is the normalization constant.

[0060] In a second aspect, a multi-objective optimization system for a loudspeaker array with sound field partition control is provided, comprising:

[0061] The first module is used to: divide the sound field area into a bright area and a dark area according to listening expectations;

[0062] The second module is used to: construct a decision variable vector according to the 0-1 type planning using the candidate speaker placement points as decision variables;

[0063] A third module is configured to construct a multi-objective optimization model for the decision variable vector, wherein the multi-objectives include a first objective based on the contrast of acoustic potential energy between bright and dark areas, a second objective based on the number of loudspeakers, and a third objective based on the normalized mean square error of the bright area and the spatial consistency of the loudspeaker signals in the bright area;

[0064] A fourth module is configured to solve the multi-objective optimization model using a multi-objective genetic algorithm to obtain a plurality of Pareto optimal solutions;

[0065] A fifth module is configured to construct a linear weighted goal programming model based on the expected values ​​of the multiple objectives, the objective function values ​​of the multiple objectives corresponding to each of the Pareto optimal solutions, and the weight coefficients of the negative deviation variables and the positive deviation variables of the objective function values ​​relative to the expected values;

[0066] The sixth module is used to solve the linear weighted objective programming model to obtain the Pareto optimal solution corresponding to the objective function value when the linear weighted objective programming model takes the minimum value, and obtain the optimal position and number of the speaker array.

[0067] The beneficial effects of the technical solution provided by this application include:

[0068] The multi-objective optimization method provided in this application utilizes a multi-objective optimization strategy that comprehensively considers the contrast between the acoustic potential energy of bright and dark areas, the normalized mean square error of the bright area, spatial consistency of the bright area, and the minimization of the number of speakers. This effectively balances sound field quality and cost, making it particularly suitable for speaker layout within the limited space of a vehicle. Furthermore, linear weighted objective programming is used to precisely select speaker positions, achieving an optimal balance between multiple objectives and further enhancing the sound field zoning effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0070] Figure 1 A flow chart of a multi-objective optimization method for a loudspeaker array with sound field zoning control provided in an embodiment of the present application;

[0071] Figure 2 A distribution diagram of the speaker array and microphone array provided in the embodiment of the present application;

[0072] Figure 3 A top view of the speaker array provided in an embodiment of the present application;

[0073] Figure 4 A side view of a speaker array provided in an embodiment of the present application;

[0074] Figure 5 The optimal speaker array diagram obtained by the multi-objective optimization method provided in the embodiment of the present application;

[0075] Figure 6 The optimal loudspeaker array diagram obtained by the single-objective optimization method provided in the comparative example of this application;

[0076] Figure 7 This is a diagram of the bright area spatial consistency results provided by the embodiment of this application;

[0077] Figure 8 This is a graph of the normalized mean square error results for the bright area provided in the embodiment of the present application;

[0078] Figure 9 This is a graph showing the contrast between the acoustic potential energy in the bright and dark areas provided in the examples of this application. DETAILED DESCRIPTION

[0079] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0080] See also Figure 1 As shown, the embodiment of the present application provides a multi-objective optimization method for a loudspeaker array with sound field partition control, which includes the following steps:

[0081] 10: Divide the sound field into bright and dark areas based on listening expectations.

[0082] Regarding step 10, when dividing the sound field area, the area where a specific sound field is desired to be generated is defined as a bright area, and the other areas are defined as dark areas.

[0083] For example, the interior of a car can be divided into four independent sound zones according to the position of the cabin, and these four cabin areas are symmetrical about the center of the roof.

[0084] Generally speaking, microphones are evenly distributed in the space where the passenger's head is located in each cabin area as control points, and speakers are arranged on the roof and doors. Therefore, candidate speaker arrangement points can be set on the roof and doors.

[0085] After the speaker array emits sound based on the input drive signal, a microphone array located above the passenger's head collects the signal. Based on the speaker drive signal and the sound signal collected by the microphone, a transfer function matrix from the speaker array to the microphone array can be derived. This matrix is ​​related to the signal frequency.

[0086] Determining the transfer function matrix specifically includes the following steps:

[0087] 101: Arrange speakers at all candidate speaker placement points, and arrange microphones in bright areas and dark areas.

[0088] 102: Give a test drive signal to each of the speakers one by one, and obtain the measured sound pressure collected by each microphone.

[0089] 103: Based on the test drive signal and the measured sound pressure, obtain the transfer function matrix G (f i ):

[0090]

[0091] Among them, p 0，mj (f i ) is the sound pressure measured by the m-th microphone at the j-th speaker arrangement candidate point, m = 1, 2, ..., M, M is the total number of microphones in the sound field area, q 0，j (f i ) is the test driving signal of the jth speaker, j=1, 2, ..., K, K is the total number of candidate points for speaker placement in the sound field area, is the transfer function from the jth loudspeaker to the mth microphone, i is the frequency point order, f i is the frequency of the i-th frequency point, i=1,2,…, f min is the lower limit of the speaker frequency, f max is the upper limit of the speaker frequency, f i+1 -f i =1Hz, f i =f min ,f min +1,f min +2,…,f max , the frequency of each frequency point increases in increments of 1 Hz.

[0092] Through step 103, different frequencies f can be obtained. i The transfer function matrix G(f i ).

[0093] 104: Based on the microphones in the bright area and the dark area, the transfer function matrix G (fi ) to obtain the transfer function matrix G from the loudspeaker to the bright area microphone B (f i ) and the transfer function matrix G from the loudspeaker to the dark area microphone D (f i ).

[0094] The sound pressure at the M microphones is a linear transformation of the driving signal in the frequency domain, and the corresponding linear transformation matrix is ​​the transfer function matrix G(f i ), from which we can get:

[0095]

[0096] Among them, P B (f i ) is the sound pressure received by the microphone in the bright area, P D (f i ) is the sound pressure received by the microphone in the dark area.

[0097] 20: Using the candidate speaker placement points as decision variables, construct a decision variable vector according to the 0-1 type planning.

[0098] Loudspeaker array optimization aims to select the optimal speaker placement from a set of candidate speaker locations. This problem can often be formulated as a combinatorial optimization problem. The control variables in this optimization model can be described using 0-1 variables. A 1 indicates that a speaker is placed at that candidate location, while a 0 indicates that a speaker is not placed at that location.

[0099] Therefore, the decision variable vector is recorded as X, which includes:

[0100] X=[x1,x2,…,x j ,…,x K ]

[0101] Among them, x j =0 or 1, when x j = 0, no speaker is set at the jth speaker placement candidate point. j =1, the jth speaker placement candidate point is used to place a speaker, j=1, 2, ..., K, where K is the total number of speaker placement candidate points in the sound field area.

[0102] Based on the common speaker positions in current vehicles, consider the door speakers and abstract the vehicle's geometric dimensions into a rectangular parallelepiped. For the sake of vehicle aesthetics, consider the speakers symmetrical about the center point of the roof. The candidate points are shown in Figure 2 A diagram showing the locations of the speaker and microphone array in a car.

[0103] The sound field area is divided into four cabin areas symmetrical about the center of the roof according to the position of the cabin in the car, so that x j′ =x j′+L =x j′+2L =x j′+3L ;

[0104] Wherein, j′=1, 2, …, L, L is the total number of candidate points for speaker placement in each cabin area, and K=4L.

[0105] 30: Construct a multi-objective optimization model with respect to the decision variable vector, wherein the multi-objectives include: a first objective based on the contrast of sound potential energy between the bright area and the dark area, a second objective based on the number of loudspeakers, and a third objective based on the normalized mean square error of the bright area and the spatial consistency of the loudspeaker signal in the bright area.

[0106] In step 30, the multi-objective optimization model is as follows:

[0107]

[0108] Among them, F1(X) is the first target based on the contrast of sound potential energy in the bright area and the dark area, F2(X) is the second target based on the number of loudspeakers, and F3(X) is the third target based on the normalized mean square error in the bright area and the spatial consistency of the loudspeaker signal in the bright area.

[0109] (1) Build the first goal

[0110] The contrast of the acoustic potential energy of the bright area and the dark area is recorded as AC i (X), and AC i (X) is:

[0111]

[0112] The AC i (X) The higher the index value, the higher the contrast between the sound potential energy in the bright area and the dark area. In the optimization design of the speaker array, it is generally hoped to obtain a better acoustic experience in the bright area by improving the contrast.

[0113] Among them, G B (f i ) is the transfer function matrix from the loudspeaker to the bright area microphone, with dimension M B ×K,M B is the number of microphones in the bright area, G B H (f i ) is G B (f i )'s conjugate transposed matrix;

[0114] G D (f i) is the transfer function matrix from the loudspeaker to the dark zone microphone, with a dimension of M D ×K,M D is the number of dark area microphones, G D H (f i ) is G D (f i )’s conjugate transposed matrix; M B +M D =M, where M is the total number of microphones in the sound field;

[0115] q i is the driving signal vector of the loudspeaker, with a dimension of K×1, where K is the total number of candidate points for loudspeaker placement within the sound field area, and q i H q i The conjugate transposed matrix of ;

[0116] q i =[x1q1(f i ),x2q2(f i ),…,x j q j (f i ),…,x K q K (f i )] T

[0117] Where j = 1, 2, ..., K, q j (f i ) is the driving signal of the jth speaker, i is the frequency point order, f i is the frequency of the i-th frequency point, i=1,2,…, f min is the lower limit of the speaker frequency, f max is the upper limit of the speaker frequency, f i+1 -f i =1Hz, f i =f min ,f min +1,f min +2,…,f max ; From the above q i As you can see from the expression, it is q i Related to the decision variable vector X.

[0118] Based on the contrast of acoustic potential energy in bright and dark areas AC i The first objective F1(X) is:

[0119]

[0120] (2) Constructing the second goal

[0121] The second goal is based on the number of speakers. When optimizing the speaker array layout, it is always hoped to reduce the number of speakers to reduce costs.

[0122] Thus, the second target F2(X) based on the number of speakers is:

[0123] F2(X)=||X||1

[0124] Here, ||·||1 is the 1-norm.

[0125] The number of speakers actually refers to the number of non-zero elements in the decision variable vector X. Therefore, the 1-norm is used as an indicator to describe it.

[0126] (3) Constructing the third goal

[0127] Regarding the error percentage of the bright area sound field, in order to overcome the difference in index comparison caused by the order of magnitude of the error, the normalized bright area sound field error ratio is used, that is, the normalized mean square error of the bright area is recorded as E r,i (X), and E r,i (X) is:

[0128]

[0129] Among them, P BT (f i ) is the target sound field matrix for bright area reconstruction, is the square of the 2-norm;

[0130] In order to describe the spatial consistency of the loudspeaker signal in the bright area, ΔL is introduced. i (X), the smaller the index is, the more concentrated the sound pressure distribution is, that is, the spatial consistency of the loudspeaker signal in the bright area is recorded as ΔL i (X), and ΔL i (X) is:

[0131] ΔL i (X) = L i,B,max (X)-L i,B,min (X)

[0132]

[0133] Among them, L i,B (X) is the sound pressure level of the microphone in the bright area, P ref (f i ) is the reference sound pressure, L i,B,max (X) is the L of all microphones in the bright area i,B The maximum value of (X), L i,B,min (X) is the L of all microphones in the bright areai,B (X) minimum value;

[0134] Therefore, the third objective F3(X) based on the normalized mean square error in the bright area and the spatial consistency of the loudspeaker signal in the bright area is:

[0135]

[0136] 40: Using a multi-objective genetic algorithm to solve the multi-objective optimization model to obtain several Pareto optimal solutions.

[0137] The multi-objective genetic algorithm (NSGA-II algorithm) is an evolutionary algorithm specifically designed to solve multi-objective optimization problems. Its main idea is to effectively search and maintain the non-dominated solution set in the optimization problem by combining the basic principles of genetic algorithms and non-dominated sorting techniques.

[0138] For multi-objective optimization models:

[0139]

[0140] Any two solutions X Δ1 and X Δ2 , if the following two conditions are met:

[0141] (1) Non-Domination: For all a = 1, 2, 3, there is F a (X Δ1 )≤F a (X Δ2 ).

[0142] (1) Improvement: There is at least one objective function F a′ (X), so that F a (X Δ1 )<F a (X Δ2 ), a′=1,2,3.

[0143] Then solve X Δ1 As the Pareto optimal solution, we can obtain a set of Pareto optimal solutions, also called the Pareto front.

[0144] 50: Construct a linear weighted goal programming model based on the expected values ​​of the multiple objectives, the objective function values ​​of the multiple objectives corresponding to each of the Pareto optimal solutions, and the weight coefficients of the negative deviation variables and the positive deviation variables of the objective function values ​​relative to the expected values.

[0145] For the three optimization objectives F a (X), give each optimization goal an expected value b a, thus, we can get the linear weighted target programming model, which is:

[0146]

[0147] Among them, F a (X)+s a -t a =b a ,a=1,2,3,X is the decision variable vector. When a=1, F1(X) is the first target based on the contrast of the sound potential energy in the bright area and the dark area. When a=2, F2(X) is the second target based on the number of loudspeakers. When a=3, F3(X) is the third target based on the normalized mean square error in the bright area and the spatial consistency of the loudspeaker signal in the bright area. a ,t a ≥0,s a ×t a =0,b a F a The expected value of (X), s a F a (X) positive deviation variable, u a is a positive deviation variable s a The weight of t a F a (X) negative deviation variable, v a is a negative deviation variable t a The weight, w a is the normalization constant.

[0148] According to F a (X)+s a -t a =b a ,a=1,2,3,F a (X) is the corresponding objective function value calculated from the Pareto optimal solution.

[0149] 60: Solve the linear weighted objective programming model to obtain a Pareto optimal solution corresponding to the objective function value when the linear weighted objective programming model takes a minimum value, and obtain the optimal position and number of the speaker array.

[0150] According to the above linear weighted goal programming model, the minimum value can be solved. At this time, F a (X) corresponds to the Pareto optimal solution, and then the decision variable vector X is obtained. According to the value of each element of the decision variable vector X (0 or 1), the optimal position and number of the speaker array can be obtained.

[0151] In order to improve the sound field partition control effect, the method provided in this application also includes optimizing the driving signal. The optimization of the driving signal specifically includes the following steps:

[0152] 301: Construct a drive signal optimization model as follows:

[0153]

[0154] 302: Solve the driving signal optimization model to obtain the driving signal vector q corresponding to the speaker with the maximum eigenvalue i * .

[0155] It is understandable that when performing step 302, due to q i The elements of are related to the decision variable vector X, that is, X=[x1,x2,…,x j ,…,x K ], x j =0 or 1, if an element is 0, correspondingly, q i The elements at the corresponding positions in are also 0. Therefore, when solving the above optimization model, when a row or a column is 0, the row or column can be temporarily removed to obtain the maximum eigenvalue. After completion, the driving signal vector is padded with 0 to obtain q i * .

[0156] This application optimizes the driving signal by constructing a driving signal optimization model to use the optimized driving signal q i * , which can further improve the contrast of sound potential energy between bright and dark areas and the spatial consistency of speaker signals in the bright area, and enhance passengers' experience of enjoying independent sound zones.

[0157] Furthermore, after completing step 302, the driving signal is optimized, further comprising the following steps:

[0158] 303: q i * After multiplying by the coefficient c, the optimization model for the coefficient c is constructed as follows:

[0159]

[0160] in, is the square of the 2-norm, P BT (f i ) is the target sound field matrix for bright area reconstruction;

[0161] 304: Solve the optimization model for coefficient c to obtain the optimal solution c for coefficient c * , change q i * Multiply by c * After that, we get c * qi * , and as the driving signal vector q of the loudspeaker i .

[0162] By constructing an optimization model for coefficient c, the optimal solution c of coefficient c is obtained. * , to apply the optimized drive signal c * q i * , which can further improve the normalized mean square error of the bright area and enhance passengers' experience of enjoying independent sound zones.

[0163] It can be seen that the multi-objective optimization method provided in the embodiment of the present application, first of all, adopts a multi-objective optimization strategy, comprehensively considering the sound potential energy contrast between bright and dark areas, the normalized mean square error of the bright area, the spatial consistency of the bright area and the minimization of the number of speakers, thereby effectively balancing the sound field quality and cost, and is particularly suitable for the speaker layout in the limited space in the car. Secondly, by optimizing the driving signal, the present application significantly improves the sound field zoning control effect, improves the sound potential energy contrast between bright and dark areas and the spatial consistency of the bright area, improves the normalized mean square error of the bright area, and enhances the passengers' enjoyment of independent sound zones. Finally, linear weighted target planning is used to accurately select the speaker position, achieve the best balance between multiple objectives, and further improve the sound field zoning effect.

[0164] In order to test the effect of the multi-objective optimization method of the present application in optimizing the layout of the loudspeaker array, the algorithm was compared with the single-objective optimization algorithm.

[0165] Table 1 Microphone location information

[0166]

[0167] Table 1 uses the left front area as an example. The ratio of 1:0.1:1.2 indicates that microphones are placed every 0.1m in the longitudinal direction of the vehicle from 1 to 1.2m; every 0.3m in the lateral direction from 0.2 to 0.8m; and every 0.2m in the height from 0.8 to 1m. Therefore, there are 18 microphones available for placement in the left front area. The same layout applies to the right front, left rear, and right rear areas.

[0168] Table 2. Location information of candidate loudspeaker placement points

[0169]

[0170] Table 2 is the location information of the candidate points for speaker layout. The specific layout of the speaker array and microphone array is as follows: the microphone array of each area is composed of 18 microphones, and the speaker array of each roof area and door area is composed of 9 speakers respectively. The specific situation of the controlled frequency points is as follows: the control frequency range is 0Hz~900Hz, and the control frequency interval is 1Hz. At the same time, considering the user's need to set the right front area (co-pilot) as a bright area and other areas as dark areas. Please refer to Table 1 for the microphone installation layout coordinates and Table 2 for the location information of the candidate points for speaker layout. For the distribution of speaker arrays and microphone arrays, please refer to Figure 2 . Figure 3 A top view of the speaker array. Figure 4 A side view of the speaker array.

[0171] In this specific case, the results of the multi-objective optimization method of this application are compared as follows:

[0172] Figure 5 This is the optimal speaker array diagram obtained by the multi-objective optimization method of this application. For comparison, Figure 6 The optimal loudspeaker array diagram of the single-objective optimization method is listed.

[0173] Figure 7 This is the result diagram of spatial consistency in the bright area. In the figure, the average consistency index of the single-objective optimization method is 5.7506dB, and the average consistency index of the multi-objective optimization method is 3.9322dB. This figure shows that the present application has better consistency.

[0174] Figure 8 This is the normalized mean square error result diagram for the bright area. In this diagram, the average value of the target sound field restoration error index of the single-objective optimization method is 0.9930, and the average value of the target sound field restoration error index of the multi-objective optimization method is 0.9945.

[0175] Figure 9 The figure shows the result of the contrast of the sound potential energy of the bright and dark areas. In the range of 0Hz to 900Hz, the average value of the contrast of the sound potential energy of the bright and dark areas of the single-objective optimization method is 13.1011dB, and the average value of the contrast of the sound potential energy of the bright and dark areas of the present application is 25.8820dB. The above results show that the present application method is superior to the single-objective optimization method in terms of the consistency of the bright area and the contrast of the sound potential energy of the bright and dark areas when the error loss of the target sound field restoration is not large.

[0176] In summary, the present application maintains a low normalized mean square error in the bright area, while improving the sound potential energy contrast between the bright and dark areas and the spatial consistency of the bright area. Therefore, the method of the present application has good loudspeaker array layout optimization performance.

[0177] Based on the above method, an embodiment of the present application further provides a multi-objective optimization system for a loudspeaker array with sound field zoning control, which includes:

[0178] The first module is used to: divide the sound field area into a bright area and a dark area according to listening expectations;

[0179] The second module is used to: construct a decision variable vector according to the 0-1 type planning using the candidate speaker placement points as decision variables;

[0180] A third module is configured to construct a multi-objective optimization model for the decision variable vector, wherein the multi-objectives include a first objective based on the contrast of acoustic potential energy between bright and dark areas, a second objective based on the number of loudspeakers, and a third objective based on the normalized mean square error of the bright area and the spatial consistency of the loudspeaker signals in the bright area;

[0181] A fourth module is configured to solve the multi-objective optimization model using a multi-objective genetic algorithm to obtain a plurality of Pareto optimal solutions;

[0182] A fifth module is configured to construct a linear weighted goal programming model based on the expected values ​​of the multiple objectives, the objective function values ​​of the multiple objectives corresponding to each of the Pareto optimal solutions, and the weight coefficients of the negative deviation variables and the positive deviation variables of the objective function values ​​relative to the expected values;

[0183] The sixth module is used to solve the linear weighted objective programming model to obtain the Pareto optimal solution corresponding to the objective function value when the linear weighted objective programming model takes the minimum value, and obtain the optimal position and number of the speaker array.

[0184] In the description of this application, it should be noted that the terms "upper" and "lower" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on this application. Unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be internal communication between two elements. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to the specific circumstances.

[0185] It should be noted that, in this application, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.

[0186] The foregoing is merely a list of specific embodiments of the present application, intended to enable those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the broadest scope consistent with the principles and novel features of the present application.

Claims

1. A multi-objective optimization method for a loudspeaker array with sound field partition control, characterized in that: It includes: According to listening expectations, the sound field area is divided into bright area and dark area; Taking the candidate speaker placement points as decision variables, the decision variable vector is constructed according to the 0-1 type planning; Constructing a multi-objective optimization model for the decision variable vector, the multi-objectives including: a first objective based on the contrast of acoustic potential energy in the bright area and the dark area, a second objective based on the number of loudspeakers, and a third objective based on the normalized mean square error in the bright area and the spatial consistency of the loudspeaker signals in the bright area; Using a multi-objective genetic algorithm to solve the multi-objective optimization model to obtain several Pareto optimal solutions; Constructing a linear weighted goal programming model based on the expected values ​​of the multiple objectives, the objective function values ​​of the multiple objectives corresponding to each of the Pareto optimal solutions, and the weight coefficients of the negative deviation variables and the positive deviation variables of the objective function values ​​relative to the expected values; The linear weighted objective programming model is solved to obtain a Pareto optimal solution corresponding to the objective function value when the linear weighted objective programming model takes a minimum value, and the optimal position and number of the loudspeaker array are obtained.

2. The multi-objective optimization method for a loudspeaker array with sound field zoning control according to claim 1, wherein: The decision variable vector is denoted as X, which includes: Among them, x j =0 or 1, when x j = 0, no speaker is set at the jth speaker placement candidate point. j =1, the jth speaker placement candidate point is used to place a speaker, j=1, 2, ..., K, where K is the total number of speaker placement candidate points in the sound field area.

3. The multi-objective optimization method for a loudspeaker array with sound field zoning control according to claim 2, wherein: The sound field area is divided into four cabin areas symmetrical about the center of the roof according to the position of the cabin in the car, so that x j′ =x j′+L =x j′+2L =x j′+3L ; Wherein, j′=1, 2, …, L, L is the total number of candidate points for speaker placement in each cabin area, and K=4L.

4. The multi-objective optimization method for a loudspeaker array with sound field zoning control according to claim 2, wherein: The multi-objective optimization model is as follows: Among them, F1(X) is the first target based on the contrast of sound potential energy in the bright area and the dark area, F2(X) is the second target based on the number of loudspeakers, and F3(X) is the third target based on the normalized mean square error in the bright area and the spatial consistency of the loudspeaker signal in the bright area.

5. The multi-objective optimization method for a loudspeaker array with sound field zoning control according to claim 4, characterized in that: The contrast of the acoustic potential energy of the bright area and the dark area is recorded as AC i (X), and AC i (X) is: Among them, G B (f i ) is the transfer function matrix from the loudspeaker to the bright area microphone, with dimension M B ×K,M B is the number of microphones in the bright area, G B H (f i ) is G B (f i )'s conjugate transposed matrix; G D (f i ) is the transfer function matrix from the loudspeaker to the dark zone microphone, with a dimension of M D ×K,M D is the number of dark area microphones, G D H (f i ) is G D (f i )’s conjugate transposed matrix; M B +M D =M, where M is the total number of microphones in the sound field; q i is the driving signal vector of the loudspeaker, with a dimension of K×1, where K is the total number of candidate points for loudspeaker placement within the sound field area, and q i H q i The conjugate transposed matrix of ; q i =[x1q1(f i ),x2q2(f i ),…,x j q j (f i ),…,x K q K (f i )] T Where j = 1, 2, ..., K, q j (f i ) is the driving signal of the jth speaker, i is the frequency point order, f i is the frequency of the ith frequency point, f min is the lower limit of the speaker frequency, f max is the upper limit of the speaker frequency, f i+1 -f i =1Hz, f i =f min ,f min +1,f min +2,…,f max ; Based on the contrast of acoustic potential energy in bright and dark areas AC i The first objective F1(X) is:

6. The multi-objective optimization method for a loudspeaker array with sound field zoning control according to claim 4, wherein: The second target F2(X) based on the number of speakers is: F2(X)=||X||1 Here, ||·||1 is the 1-norm.

7. The multi-objective optimization method for a loudspeaker array with sound field zoning control according to claim 5, wherein: The normalized mean square error of the bright area is recorded as E r,i (X), and E r,i (X) is: Among them, P BT (f i ) is the target sound field matrix for bright area reconstruction, is the square of the 2-norm; The spatial consistency of the loudspeaker signal in the bright area is recorded as ΔL i (X), and ΔL i (X) is: ΔL i (X)=L i,B,max (X)-L i,B,min (X) Among them, L i,B (X) is the sound pressure level of the microphone in the bright area, P ref (f i ) is the reference sound pressure, L i,B,max (X) is the L of all microphones in the bright area i,B The maximum value of (X), L i,B,min (X) is the L of all microphones in the bright area i,B (X) minimum value; The third objective F3(X) based on the normalized mean square error in the bright area and the spatial consistency of the loudspeaker signal in the bright area is:

8. The multi-objective optimization method for a loudspeaker array with sound field zoning control according to claim 5, wherein: The method further includes determining a transfer function matrix, wherein determining the transfer function matrix specifically includes: Place speakers at all candidate speaker placement points, and place microphones in bright and dark areas; giving a test drive signal to each of the speakers one by one, and obtaining a measured sound pressure collected by each microphone; Based on the test drive signal and the measured sound pressure, the transfer function matrix G(f i ): Among them, p 0，mj (f i ) is the sound pressure measured by the m-th microphone at the j-th speaker placement candidate point, m = 1, 2, ..., M, q 0，j (f i ) is the test driving signal of the jth speaker, is the transfer function from the jth loudspeaker to the mth microphone; According to the microphones in the bright area and the dark area, the transfer function matrix G(f i ) to obtain the transfer function matrix G from the loudspeaker to the bright area microphone B (f i ) and the transfer function matrix G from the loudspeaker to the dark area microphone D (f i ).

9. The multi-objective optimization method for a loudspeaker array with sound field zoning control according to claim 5, wherein: The method further includes optimizing the driving signal, and the optimization of the driving signal specifically includes: Construct the drive signal optimization model as follows: Solve the driving signal optimization model to obtain the driving signal vector q corresponding to the speaker with the maximum eigenvalue i * .

10. The multi-objective optimization method for a loudspeaker array with sound field zoning control according to claim 9, wherein: Optimizing the drive signal also includes: q i * After multiplying by the coefficient c, the optimization model for the coefficient c is constructed as follows: in, is the square of the 2-norm, P BT (f i ) is the target sound field matrix for bright area reconstruction; Solve the optimization model about coefficient c to obtain the optimal solution c of coefficient c * , change q i * Multiply by c * After that, we get c * q i * , and as the driving signal vector q of the loudspeaker i .

11. The multi-objective optimization method for a loudspeaker array with sound field zoning control according to claim 1, wherein: The linear weighted goal programming model is: Among them, F a (X)+s a -t a =b a ,a=1,2,3,X is the decision variable vector. When a=1, F1(X) is the first target based on the contrast of the sound potential energy in the bright area and the dark area. When a=2, F2(X) is the second target based on the number of loudspeakers. When a=3, F3(X) is the third target based on the normalized mean square error in the bright area and the spatial consistency of the loudspeaker signal in the bright area. a ,t a ≥0,s a ×t a =0,b a F a The expected value of (X), s a F a (X) positive deviation variable, u a is a positive deviation variable s a The weight of t a F a (X) negative deviation variable, v a is a negative deviation variable t a The weight, w a is the normalization constant.

12. A multi-objective optimization system for a loudspeaker array with sound field zoning control, characterized in that: It includes: The first module is used to: divide the sound field area into a bright area and a dark area according to listening expectations; The second module is used to: construct a decision variable vector according to the 0-1 type planning using the candidate speaker placement points as decision variables; A third module is configured to construct a multi-objective optimization model for the decision variable vector, wherein the multi-objectives include a first objective based on the contrast of acoustic potential energy between bright and dark areas, a second objective based on the number of loudspeakers, and a third objective based on the normalized mean square error of the bright area and the spatial consistency of the loudspeaker signals in the bright area; A fourth module is configured to solve the multi-objective optimization model using a multi-objective genetic algorithm to obtain a plurality of Pareto optimal solutions; A fifth module is configured to construct a linear weighted goal programming model based on the expected values ​​of the multiple objectives, the objective function values ​​of the multiple objectives corresponding to each of the Pareto optimal solutions, and the weight coefficients of the negative deviation variables and the positive deviation variables of the objective function values ​​relative to the expected values; The sixth module is used to solve the linear weighted objective programming model to obtain the Pareto optimal solution corresponding to the objective function value when the linear weighted objective programming model takes the minimum value, and obtain the optimal position and number of the speaker array.

Images