Multi-target positioning identification method based on array signals
阅读说明：本技术 一种基于阵列信号的多目标定位识别方法 (Multi-target positioning identification method based on array signals ) 是由 理华 孟晓辉 王耀辉 于 2018-09-11 设计创作，主要内容包括：本发明公开了一种基于阵列信号的多目标定位识别方法,所述方法包括：一种基于阵列信号的多目标定位识别方法,所述方法包括：步骤1)利用阵列盲信号处理将N个独立信源从混叠信号中分离出来；步骤2)对多通道的阵列信号进行频带分解,分离出M个不同频带的信号；将空间平面分为P*Q个网格,P为网格的行总数,Q为网格的列总数；步骤3)利用阵列信号处理的声源定位算法分别计算M个频带在每个网格位置上输出的功率；步骤4)基于步骤3)的每个网格的M个频带的功率,将每一个网格的信号恢复到时域,获得每个网格上的时域信号；步骤5)将步骤1)的每个独立信源分别与P*Q个时域信号进行匹配,匹配到的网格位置为独立信源的位置,共匹配P*Q*N次。(The invention discloses a multi-target positioning identification method based on array signals, which comprises the following steps: a multi-target positioning identification method based on array signals comprises the following steps: step 1) separating N independent information sources from aliasing signals by array blind signal processing; step 2) carrying out frequency band decomposition on the multi-channel array signal to separate out M signals with different frequency bands; dividing the space plane into P × Q grids, wherein P is the total number of rows of the grids, and Q is the total number of columns of the grids; step 3) respectively calculating the output power of the M frequency bands at each grid position by using a sound source positioning algorithm of array signal processing; step 4) recovering the signal of each grid to the time domain based on the power of the M frequency bands of each grid in the step 3), and obtaining a time domain signal on each grid; and 5) respectively matching each independent signal source in the step 1) with P x Q time domain signals, wherein the matched grid position is the position of the independent signal source and is matched for P x Q x N times.)
1. A multi-target positioning identification method based on array signals comprises the following steps:
step 1) separating N independent information sources from aliasing signals by array blind signal processing;
step 2) carrying out frequency band decomposition on the multi-channel array signal to separate out M signals with different frequency bands; dividing the space plane into P × Q grids, wherein P is the total number of rows of the grids, and Q is the total number of columns of the grids;
step 3) respectively calculating the output power of the M frequency bands at each grid position by using a sound source positioning algorithm of array signal processing;
step 4) recovering the signal of each grid to the time domain based on the power of the M frequency bands of each grid in the step 3), and obtaining a time domain signal on each grid;
and 5) respectively matching each independent signal source in the step 1) with P x Q time domain signals, wherein the matched grid position is the position of the independent signal source and is matched for P x Q x N times.
2. The method for identifying multiple target locations in an array signal according to claim 1, wherein the step 3) is implemented by the following steps:
calculating the power P of the M-th frequency band output at each grid position by using the SRP-PHAT method, wherein M is more than or equal to 1 and less than or equal to Mm(s)：
Wherein L is the number of channels of the array signal, Xk(ω) is the k channel signal Xk(t) windowed Fourier transform, τkPointing the controllable delay at the grid (p, q) for the kth channel;
the M bands output P x Q x M powers on P x Q grids.
3. The method for identifying multiple target locations in an array signal according to claim 2, wherein the step 4) is implemented by the following steps: the time domain signal x (m, s) at grid (p, q) is:
where M is 0, 1, 2.. M-1, M denotes a time point of the time-domain signal,
4. the method for identifying multiple target locations in an array signal according to claim 3, wherein the step 5) is implemented by the following steps:
calculating a correlation coefficient using a cross-correlation method for each independent source X obtained in step 1) and a time-domain signal Y (p, q) at a grid (p, q) obtained in step 4):
wherein Cov (X, Y (p, q)) is the covariance of X and Y (p, q), Var [ X ] is the variance of X, and Var [ Y (p, q) ] is the variance of Y (p, q);
the grid with the largest correlation coefficient is:
Technical Field
The invention relates to the field of voice signal processing, in particular to a multi-target positioning identification method based on array signals.
Background
The microphone array signal processing technology is derived from systems such as radar and sonar. Microphone array sound source localization has wide application value in many fields such as communication, mobile robots and hearing aid devices. In these applications, it is essential to estimate the sound source position, and both the positioning accuracy and the algorithm real-time performance need to be considered. The conventional DOA estimation method, such as the controllable power response sound source localization method (SRP-PHAT), can only estimate the position of the sound source, but cannot distinguish the position of the sound source with similar frequency, i.e., cannot establish a one-to-one relationship between the sound source and the position.
Disclosure of Invention
The invention aims to overcome the technical defects and provides a multi-target positioning identification method based on array signals, which can establish a one-to-one correspondence relationship between independent information sources or sparse information sources with similar frequencies and the sound source positions of the independent information sources or the sparse information sources, thereby realizing the positioning of the sound source and the type positioning of the sound source.
In order to achieve the above object, the present invention provides a multi-target positioning identification method based on array signals, the method comprising:
step 1) separating N independent information sources from aliasing signals by array blind signal processing;
step 2) carrying out frequency band decomposition on the multi-channel array signal to separate out M signals with different frequency bands; dividing the space plane into P × Q grids, wherein P is the total number of rows of the grids, and Q is the total number of columns of the grids;
step 3) respectively calculating the output power of the M frequency bands at each grid position by using a sound source positioning algorithm of array signal processing;
step 4) recovering the signal of each grid to the time domain based on the power of the M frequency bands of each grid in the step 3), and obtaining a time domain signal on each grid;
and 5) respectively matching each independent signal source in the step 1) with P x Q time domain signals, wherein the matched grid position is the position of the independent signal source and is matched for P x Q x N times.
As an improvement of the above method, the specific implementation process of step 3) is as follows:
by using SThe RP-PHAT method calculates the power P of the M-th frequency band which is output at each grid position and is more than or equal to 1 and less than or equal to Mm(s)：
Wherein L is the number of channels of the array signal, Xk(ω) is the k channel signal Xk(t) windowed Fourier transform, τkPointing the controllable delay at the grid (p, q) for the kth channel;
the M bands output P x Q x M powers on P x Q grids.
As an improvement of the above method, the specific implementation process of step 4) is as follows: the time domain signal x (m, s) at grid (p, q) is:
where M is 0, 1, 2.. M-1, M denotes a time point of the time-domain signal,
as an improvement of the above method, the specific implementation process of step 5) is:
calculating a correlation coefficient using a cross-correlation method for each independent source X obtained in step 1) and a time-domain signal Y (p, q) at a grid (p, q) obtained in step 4):
wherein Cov (X, Y (p, q)) is the covariance of X and Y (p, q), Var [ X ] is the variance of X, and Var [ Y (p, q) ] is the variance of Y (p, q);
the grid with the largest correlation coefficient is:
The invention has the advantages that:
1. according to the invention, firstly, a blind source separation method is utilized to restore the independent sound sources or the sparse sound sources which are mixed together, so that the time domain and frequency domain characteristics of a plurality of sound sources can be obtained, and the type distinction of the sound sources can be realized;
2. the method of the invention uses the SRP algorithm of multiple frequency bands to position the sound source according to different frequencies, and uses the characteristics of different sound sources to match and finally establishes the one-to-one corresponding relation between the sound source and the position.
Drawings
FIG. 1 is a schematic diagram of the calculation of spectral characteristics of different sources according to the present invention;
FIG. 2 is a schematic diagram illustrating the calculation of spectral characteristics at different locations according to the present invention;
fig. 3 is a schematic diagram of the present invention for determining the location of different sources using a matching algorithm.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments.
The invention provides a multi-target positioning identification method based on array signals, which comprises the following steps:
step 1) processing N blind signals by using an arrayIndependent information source s1(t)，…sN(t) separating from the aliased signal;
iterative computation is performed by adopting a natural gradient method, and the algorithm makes the following assumptions: signal s with N independent signal sources1(t)，…sN(t) and observed quantities x for L independent channels1(t)，…xL(t) (known), the observed quantity and the signal source have the following relationship:
x(t)＝As(t)
wherein x (t) ═ x1(t)，…xL(t)]T，s(t)＝[s1(t)，…sN(t)]TAnd A is an NxL coefficient matrix, the original problem becomes the independence of the known x (t) and s (t), and the estimation problem of s (t) is solved. The following formula is assumed:
y(t)＝Wx(t)
where y (t) is the estimate of s (t) and W is an L N coefficient matrix, the problem becomes how effectively the matrix W is estimated. The iterative process of estimation is:
1) initialization: w (0) is an identity matrix;
2) the following steps are executed in a loop until the difference between W (n +1) and W (n) is less than a prescribed value epsilon (the method for calculating the matrix difference may be set), or the number of iterations may be prescribed.
3) Using formulas
y (n) ═ w (n) y (n-1), where y (-1) ═ x
4) The following calculation formula is used:
W(n+1)＝W(n)+η(n)[I-φ(y(n))yT(n)]W(n)
where w (n) is the matrix to be estimated, η (n) is the step size, and phi (y) is a non-linear transformation, such as phi (y) being phi (y) or phi (y)3) In actual calculation, y is an L multiplied by k matrix, and k is the number of sampling points.
5) W (n) after convergence is estimated W
Obtaining signals s of N independent information sources by using y (t) ═ Wx (t)1(t)，…sN(t)。
Step 2) performing frequency band decomposition on the multichannel array signals by using Fast Fourier Transform (FFT), and separating signals of different frequency bands; assume that there are M bands: m1, M2, … MM, respectively; dividing the space plane into P × Q grids, wherein P is the total number of rows of the grids, and Q is the total number of columns of the grids;
step 3) respectively calculating the output power of the M frequency bands at each grid position by using a PHAT algorithm for array signal processing;
calculating the power P of the M-th frequency band output at each grid position by using the SRP-PHAT method, wherein M is more than or equal to 1 and less than or equal to Mm(s)：
Wherein L is the number of channels of the array signal, Xk(ω) is the k channel signal Xk(t) windowed Fourier transform, τkPointing the controllable delay at the grid (p, q) for the kth channel;
the M bands output P x Q x M powers on P x Q grids.
After the SRP-PHAT algorithm, the output power of each grid can be obtained; thus, P × Q groups of data are shared, and M bands will have P × Q M data;
step 4) restoring the signal of each grid by using inverse fast Fourier transform (ifft) to restore the signal to a time domain; sharing P x Q group waveforms; as shown in fig. 2;
in a certain grid, M data exist on M frequency bands, and by using the data, the data in each grid are restored to a time domain by using inverse Fourier transform, so that P × Q time domain signals are obtained.
The time domain signal x (m, s) at grid (p, q) is:
where M is 0, 1, 2.. M-1, M denotes a time point of the time-domain signal,
step 5) matching each individual source X of step 1) with P × Q time domain signals Y (P, Q): the correlation coefficient was calculated using the cross-correlation method:
wherein Y (p, q) ═ X (m, s), Cov (X, Y) is X, the covariance of Y, Var [ X ] is the variance of X, and Var [ Y (p, q) ] is the variance of Y (p, q);
the grid with the highest correlation coefficient is:
The process needs to be carried out for N times aiming at N independent information sources, and then the positions of the N independent information sources on a P × Q grid can be determined; as shown in fig. 3.
Finally, it should be noted that the above embodiments are only used for illustrating the technical solutions of the present invention and are not limited. Although the present invention has been described in detail with reference to the embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted without departing from the spirit and scope of the invention as defined in the appended claims.
- 上一篇：一种医用注射器针头装配设备
- 下一篇：车载雷达控制方法、装置、设备和存储介质