Signal source positioning method based on 3D-MUSIC-Like

Abstract

The invention discloses a signal source positioning method based on 3D-MUSIC-Like, and relates to the field of signal source positioning. Different from a traditional MUSIC-Like algorithm which carries out high-order cumulant processing on a one-dimensional data set, the method carries out high-order reconstruction on three-dimensional data collected by a three-dimensional microphone array, and carries out spatial search through a spectrum peak function containing a three-dimensional high-order so as to determine the position of a signal source. According to the invention, the signal source positioning range is greatly improved, the precision and the anti-interference capability are enhanced, and all-weather signal early warning and information uploading of the to-be-detected area can be realized.

Description

Technical Field

[0001] The present invention relates to the field of signal source positioning, and in particular to a vehicle-mounted microphone 3D-MUSIC-LIKE algorithm based on subspace decomposition and its intelligent detection application. Background Art

[0002] The MUSIC algorithm is a super-resolution algorithm and one of the more important and classic subspace algorithms. Currently, the optimization based on the MUSIC algorithm is divided into two categories: one is to optimize the accuracy of the algorithm; the other is to optimize the efficiency of the algorithm, and most of them are based on one-dimensional array elements.

[0003] Generally speaking, the traditional one-dimensional MUSIC algorithm needs to use the retrieval function to perform a global search of the entire space of interest when locating the signal source. Due to the orthogonality of the noise subspace and the signal angle matrix, the maximum point of the function is the signal source orientation; but the prerequisite of this theory is that the noise interference in the signal source is low. When the signal source is not pure enough or the coherence is high, the noise subspace and the signal angle matrix are no longer orthogonal, that is, the retrieval formula is invalid. At the same time, since the retrieval range of the one-dimensional microphone array is small, it will also lead to inaccurate positioning accuracy. Therefore, the research on large-scale signal source filtering under the high-dimensional microphone array element structure is particularly important. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a signal source positioning method based on 3D-MUSIC-Like, comprising the following steps:

[0005] S1. Construct a uniform cubic array so that it can receive audio signals in the form of a three-dimensional array and construct a three-dimensional array model;

[0006] S2, inserting a fourth-order moment suitable for the three-dimensional array element into the signal matrix, and calculating a fourth-order cumulant matrix, performing singular value decomposition on the fourth-order cumulant matrix, and decomposing a three-dimensional high-order noise subspace and a signal subspace;

[0007] S3, using a three-dimensional high-order spectrum peak search function to retrieve the signal source position;

[0008] S4, placing the microphone array embedded with the 3D-MUSIC-Like algorithm on an intelligent tracking vehicle, and planning the detection range with black segments;

[0009] S5. Divide the planned detection road section into multiple equal sections, and use the center point of each section as the detection stop point, that is, the detection car briefly stops at the store location to run the above 3D-MUSIC-Like algorithm.

[0010] Furthermore, the parameter configuration of the three-dimensional array includes:

[0011] In a one-dimensional linear array, the steering vector of the one-dimensional linear array is:

[0012]

[0013] The first linear array on the x-axis in the plane is equivalent to the reference array element, and each subsequent linear array is regarded as the offset of the first reference linear array in the y-axis direction, and the two-dimensional steering vector is obtained as:

[0014]

[0015] The two-dimensional linear array on the bottom surface, i.e., the xy-axis plane, is regarded as the reference sub-array, and each subsequent two-dimensional array is regarded as the offset array of the reference sub-array on the z-axis. The steering vector of the three-dimensional array is obtained as:

[0016]

[0017] Among them, A xy is the steering vector of the reference two-dimensional subarray, D is A z The diagonal matrix constructed by the rows corresponding to the matrix, A z is the guidance vector on the z-axis, as follows:

[0018]

[0019] Furthermore, the fourth-order cumulant matrix is:

[0020]

[0021] Furthermore, the spectrum peak search function is P(ii,jj)=1 / abs(b'*Gn*b).

[0022] Furthermore, a three-dimensional microphone array is installed in an intelligent patrol car, and the above-mentioned 3D-MUSIC-Like algorithm is imported into the microphone array, and an intelligent patrol integrated device is formed with the intelligent patrol car. Corresponding black strips are laid in the area to be detected as the patrol route of the intelligent patrol car, and the route is appropriately divided into equal parts, and the entire macro area is divided into several micro detection areas. The middle position of each micro detection area is used as the detection retention point, and the time of each movement of the patrol car is set according to the position of the detection retention point, that is, the car moves a specified distance at a certain interval and stops at the detection retention point of each micro detection area each time;

[0023] After the inspection is completed, 3D-MUSIC-Like monitoring data with the same number of detection retention points will be obtained. The data group will be uploaded to the detection terminal. After summary and analysis by the staff, the location of the signal source will be determined, thereby realizing the positioning of the signal within a large range.

[0024] Compared with the prior art, the present invention has the following technical effects:

[0025] The present invention is based on the algorithm optimization of the three-dimensional cubic microphone array. Compared with the low-dimensional array, the three-dimensional microphone array itself has a significant advantage in the signal source positioning range, but in the case of noise interference, especially the Gaussian white noise, its accuracy is seriously affected. In order to make up for this defect, the MUSIC-Like algorithm is innovatively applied to the three-dimensional microphone array. The present invention can greatly improve the signal source positioning range, and the accuracy and anti-interference ability are also enhanced; at the same time, thanks to the cooperation of the inspection car, it can realize all-weather unmanned detection and information upload within the range, and together with other intelligent detection equipment, it forms an intelligent detection system.

[0026] The concept, specific structure and technical effects of the present invention will be further described below in conjunction with the accompanying drawings to fully understand the purpose, characteristics and effects of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 is a three-dimensional model received according to array data of the present invention;

[0028] Figure 2 This is the simulation positioning result of the traditional 3D MUSIC algorithm under interference;

[0029] Figure 3 It is the simulation positioning result of the vehicle microphone 3D-MUSIC-LIKE algorithm according to the present invention under the same interference. DETAILED DESCRIPTION

[0030] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0031] It should be noted that the illustrations provided in the following embodiments are only used to illustrate the basic concept of the present invention in a schematic manner, and thus the illustrations only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0032] While some exemplary embodiments of the present invention have been described for the purpose of illustration, it should be understood that the present invention may be implemented in other ways not specifically shown in the drawings.

[0033] In a specific embodiment, a signal source positioning method based on 3D-MUSIC-Like is provided, comprising the following steps:

[0034] S1. Fix a plurality of microphone sensors on a plastic bracket, set a three-dimensional steering vector matrix for the three-dimensional microphone array, and construct a signal model using a specific vector matrix;

[0035] A two-dimensional array is formed by several linear arrays, and then a uniform cubic three-dimensional array element is formed by stacking several surface arrays, such as Figure 1 As shown in the figure, the three-dimensional cube array is used as the data receiving model, and the signal source is the far-field model. The key to the three-dimensional array's role in signal source positioning lies in the selection of reference array elements and the setting of the array steering vector. The specific parameter configuration of the three-dimensional array includes:

[0036] In a one-dimensional linear array, the first array element is generally set as the reference array element, and then the delay parameter of the signal transmission is given according to the set array element spacing, thereby forming the steering vector of the one-dimensional linear array, as shown in the following formula:

[0037]

[0038] The two-dimensional linear array usually takes the first linear array on the x-axis in the plane as the reference array element, and each subsequent linear array is regarded as the offset of the first reference linear array in the y-axis direction. Therefore, the two-dimensional steering vector is as follows:

[0039]

[0040] The three-dimensional cubic array element uses this idea unchanged, and regards the two-dimensional linear array on the bottom surface, that is, the xy-axis plane, as the reference sub-array. Each subsequent two-dimensional array is regarded as an offset array of the reference sub-array on the z-axis. Therefore, the steering vector on the z-axis needs to be set first, as shown in the following formula:

[0041]

[0042] So the steering vector of the three-dimensional array is as follows:

[0043]

[0044] Where Axy is the steering vector of the reference two-dimensional sub-array, and D is the diagonal matrix constructed by the rows corresponding to the Az matrix, from which the complete three-dimensional array element steering vector is derived.

[0045] S2, inserting a fourth-order moment suitable for the three-dimensional array element into the signal matrix, and calculating a fourth-order cumulant matrix, performing singular value decomposition on the fourth-order cumulant matrix, and decomposing the noise subspace and the signal subspace;

[0046] After having the guiding matrix of the three-dimensional array, it is possible to construct the actual signal source data matrix for alignment analysis and retrieval. In the traditional MUSIC algorithm, the covariance matrix of the data is constructed and the eigenvalue decomposition is performed on it. The eigenvectors with larger eigenvalues ​​are expanded into the signal subspace, and the eigenvectors with smaller eigenvalues ​​are expanded into the noise subspace. In the 3D-MUSIC-LIKE algorithm for vehicle-mounted microphones, the fourth-order cumulant matrix C is first constructed using the number and structure of the array elements of the three-dimensional array. Since the fourth-order cumulant is insensitive to Gaussian signals, the constructed fourth-order cumulant matrix not only enables the three-dimensional array elements to obtain more information, but also greatly enhances its ability to resist noise interference. The fourth-order cumulant matrix is ​​then subjected to singular value decomposition to obtain the formula C=BC s B, and then import the data matrix of the three-dimensional array element into it, and then use the fourth-order cumulant to span the noise and signal subspaces according to the size of the singular values.

[0047] Assume that multiple independent non-Gaussian far-field narrowband signals are incident on a three-dimensional cubic array composed of M omnidirectional array elements. The spatial signals are independent of each other, and the signal and noise are also statistically independent. The noise obeys Gaussian distribution. Then the matrix form of the array receiving signal is X(t) = A(θ)S(t) + N(t). Construct a fourth-order cumulant and decompose it into the form of C = BC S B, where:

[0048]

[0049] Pair Matrix C S The corresponding three-dimensional fourth-order noise and signal subspaces are delineated according to the size of the eigenvalues. The eigenvectors corresponding to the P larger eigenvalues ​​form a three-dimensional fourth-order signal subspace, and the remaining smaller eigenvalues ​​form a three-dimensional fourth-order noise subspace. The data-steering vector of the three-dimensional array element has an orthogonal relationship with the fourth-order noise matrix, thus obtaining the 3D-MUSIC-LIKE retrieval algorithm for vehicle-mounted microphones.

[0050] S3. Use the three-dimensional high-order spectrum peak search function to retrieve the signal source position.

[0051] After decomposing the fourth-order cumulant matrix into signal and noise subspaces, the property that the signal direction matrix of the three-dimensional array element is orthogonal to its noise subspace is used to construct the 3D-MUSIC-LIKE retrieval function of the vehicle microphone P(ii,jj)=1 / abs(b'*Gn*b), where b is the noise subspace separated from the fourth-order cumulant matrix under the three-dimensional guidance matrix, and Gn is the product of the signal space matrix and its own conjugate transpose. Due to the orthogonality of the signal and noise spaces, the coordinates of the maximum value after the function retrieval is completed are the positions of the signal sources.

[0052] S4. Place the microphone array embedded with the 3D-MUSIC-Like algorithm on an intelligent tracking vehicle, and use black segments to plan the detection range.

[0053] The three-dimensional microphone array is installed into the intelligent patrol car, and the above-mentioned 3D-MUSIC-Like algorithm is imported into the microphone array to form an integrated intelligent patrol device with the intelligent patrol car. Corresponding black strips are laid in the area to be inspected as the inspection route of the intelligent patrol car.

[0054] S5. Divide the planned inspection road section into multiple equal sections, and use the center point of each section as the inspection stop point, that is, the inspection car briefly stops at the store location to run the above 3D-MUSIC-Like algorithm. Divide the entire macro area into several micro inspection areas, and use the middle position of each micro inspection area as the inspection retention point. Set the time of each movement of the inspection car according to the position of the detection retention point, that is, the car moves a specified distance at a certain interval and stops at the detection retention point of each micro inspection area each time. Combine the intelligent patrol car with the three-dimensional microphone array, and use the patrol line movement function of the car to enable the microphone array to detect a wider area and realize all-weather unmanned detection and early warning.

[0055] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A signal source localization method based on 3D-MUSIC-Like, characterized in that: The steps include: S1. Construct a uniform cubic array so that it can receive audio signals in the form of a three-dimensional array and construct a three-dimensional array model; S2, inserting a fourth-order moment suitable for the three-dimensional array element into the signal matrix, and calculating a fourth-order cumulant matrix, performing singular value decomposition on the fourth-order cumulant matrix, and decomposing a three-dimensional high-order noise subspace and a signal subspace; S3, using a three-dimensional high-order spectrum peak search function to retrieve the signal source position; S4, placing the microphone array embedded with the 3D-MUSIC-Like algorithm on an intelligent tracking vehicle, and planning the detection range with black segments; S5. Divide the planned detection road section into multiple equal sections, and use the center point of each section as the detection stop point, that is, the detection car briefly stops at the store location to run the above 3D-MUSIC-Like algorithm.

2. The signal source localization method based on 3D-MUSIC-Like according to claim 1, characterized in that: The parameter configuration method of the three-dimensional array includes: In a one-dimensional linear array, the steering vector of the one-dimensional linear array is: The first linear array on the x-axis in the plane is equivalent to the reference array element, and each subsequent linear array is regarded as the offset of the first reference linear array in the y-axis direction, and the two-dimensional steering vector is obtained as: The two-dimensional linear array on the bottom surface, i.e., the xy-axis plane, is regarded as the reference sub-array, and each subsequent two-dimensional array is regarded as the offset array of the reference sub-array on the z-axis. The steering vector of the three-dimensional array is obtained as: Among them, A xy is the steering vector of the reference two-dimensional subarray, D is A z The diagonal matrix constructed by the rows corresponding to the matrix, A z is the guidance vector on the z-axis, as follows:

3. The signal source localization method based on 3D-MUSIC-Like according to claim 1, characterized in that: The three-dimensional fourth-order cumulant matrix is: where x k (t) is a three-dimensional data model, which uses multiple Kronecker product processing to construct a three-dimensional fourth-order quantity.

4. The signal source localization method based on 3D-MUSIC-Like according to claim 1, characterized in that: The spectrum peak search function is P(ii,jj)=1 / abs(b'*Gn*b); where b is a vector containing three-dimensional high-order information, and spatial retrieval is performed using the orthogonality with its noise subspace, thereby achieving accurate positioning of targets within a large range.

Images