KR20200059574A - Method for processing the signal for an adaptive beamformer using sub-band steering covariance matrix - Google Patents

Abstract

The present invention relates to a signal processing method of an adaptive beamformer. More specifically, when calculating the beam output using sensor data in an underwater detection device such as a sonar, by calculating narrowband and wideband beam outputs using the sub-band steering covariance matrix, a problem of singular matrix occurrence due to frequency dependency elimination and inverse matrix operation is solved so that the orientation resolution of the adaptive beamformer can be improved.

Description

Method of processing the signal for an adaptive beamformer using sub-band steering covariance matrix}

The present invention relates to a signal processing method of an adaptive beamformer, and more specifically, when calculating the beam output using sensor data in an underwater detection device such as a sonar, narrowband and wideband using a subband steering covariance matrix By calculating the beam output, it relates to a signal processing method capable of improving the azimuth resolution of the adaptive beamformer by removing the frequency dependency and solving the problem of generating a singular matrix due to inverse matrix calculation.

Generally, SONAR (Sound Navigation And Ranging) uses sound waves to detect the position of a target in the water, and detects noise emitted from a target present in the water, or transmits a sound wave pulse to be located at a certain distance. It functions to detect the target by receiving and analyzing the signal reflected from the target and returning.

These sonars are used for estimating and detecting a target's orientation by performing beamforming techniques with sound waves received from multiple acoustic sensors from a continuous time.

The above beamforming is an implementation method in which energy received from an antenna is concentrated along a specific direction, and communication by minimizing the effect of signal interference by radiating a directional beam in a corresponding direction instead of an existing antenna using an omnidirectional beam It is a technology to increase quality and system channel capacity.

The sonar using this beamforming can maintain the best detection performance when the underwater target is stationary, but in reality, most of the underwater targets, which are the targets of the sonar detection, are detected while moving, so the sonar detection performance is best maintained. It becomes difficult.

That is, when the target moves, the size of the beamforming output signal of the sonar decreases and the energy of the target spreads to the beam of the adjacent direction, resulting in deterioration of detection and position tracking performance and difficulty in separating and detecting adjacent targets.

In general, the sonar system, as shown in Figure 1, a plurality of acoustic sensors (array) 10, the signal receiving processor 11, so as to be able to detect the orientation, distance, depth, etc. of the target from the received signal, It consists of a beamformer 12 and a detector 13.

At this time, since the horizontal distance between the underwater target and each acoustic sensor 10 is different, a time difference corresponding to the difference of the traveling distance occurs in the signal of the received target. Then, the signal emitted from the target is received at a time difference in the same waveform to each signal receiving processor 11, and the beamformer 12 performs beamforming to calculate and correct the arrival time difference of the signals received by the sensor, thereby outputting the sensor. Maximizing the sum of power, the detector 13 detects the target's bearing, distance, and depth.

FIG. 2 is a conceptual diagram for beamforming in such a sonar system, in which beamforming of the sonar system receives target signals at θs azimuth from each acoustic sensor 21 to extract frequency components (x (ω), 22). Beam summing is performed by deriving the sum 24 by adding an appropriate weighting value 23 to each sensor signal.

When performing such beamforming, a homogeneous frequency beamforming technique is used. That is, as a technique of obtaining the average value by collecting the same frequency components by obtaining the covariance matrix of each signal fragment, for example, extracting the components corresponding to the same frequency ω0 from the N signal fragments received by the M-th sensor, and then covariance matrix It is to estimate the beamforming.

The covariance is a value indicating the degree of correlation between two random variables, and is a concept different from the variance indicating the degree of dispersion of one variable. If the value of one of the two variables tends to rise, and the other value is also correlated with the tendency to rise, the value of the covariance becomes positive, and conversely, the value of one of the two variables tends to increase. When the other values tend to fall, the value of the covariance becomes negative. This covariance is a value to understand the tendency of the correlation to rise or fall.

On the other hand, as the beamformer, an STMV (steered minimum variance) beamformer is known as a type of adaptive beamformer. The STMV beamformer steers the covariance matrix to eliminate the frequency dependency by steering the covariance matrix so as to solve the problem that a singular matrix occurs when performing the inverse matrix operation of the covariance matrix using a signal fragment, and averages the covariance matrices for all frequencies. It is a beamformer that solves the singular matrix problem when calculating the matrix inverse matrix.

However, such a STMV beamformer is also continuously required a method to further improve the orientation resolution by further reducing the frequency dependence.

The present invention was devised to solve the above-mentioned problems as described above, and the signal processing method of the adaptive beamformer using a subband steering covariance matrix capable of improving the orientation resolution of the STMV beamformer, an adaptive beamformer. Gives

The configuration of the present invention for achieving the above technical problem is to calculate a narrowband and a wideband beam output using a subband steering covariance matrix, thereby eliminating a frequency dependency and solving a singular matrix generation problem due to inverse matrix calculation. Characterized by the configuration.

The signal processing method of the adaptive beamformer using the subband steering covariance matrix of the present invention having the above configuration provides a configuration for calculating narrowband and wideband beam output using the subband steering covariance matrix, thereby eliminating frequency dependency. And solving the problem of generating a singular matrix due to inverse matrix calculation, thereby improving the azimuth resolution of the adaptive beamformer.

1 is a block diagram of a typical sonar system.
2 is a conceptual diagram of beamforming of a sonar system.
3 is a flowchart of a signal processing method of the adaptive beamformer of the present invention.

Hereinafter, the signal processing method of the adaptive beamformer using the subband steering covariance matrix of the present invention will be described in detail based on the accompanying drawings.

However, the disclosed drawings are provided as examples for sufficiently conveying the spirit of the present invention to those skilled in the art. Therefore, the present invention is not limited to the drawings presented below and may be embodied in other aspects.

In addition, unless otherwise defined in terms used in the specification of the present invention, those of ordinary skill in the art to which the present invention pertains have the meanings commonly understood, and the subject matter of the present invention in the following description and the accompanying drawings Detailed descriptions of well-known functions and configurations that may be unnecessarily obscured are omitted.

The signal processing method of the adaptive beamformer using the subband steering covariance matrix of the present invention, unlike using the entire frequency band when obtaining the weight vector of the conventional STMV beamformer, estimates the steering covariance matrix for each subband frequency and negotiates By calculating the band and broadband beam output, the frequency dependence is reduced to improve the azimuth resolution.

Hereinafter, the signal processing method of the adaptive beamformer of the present invention will be described in detail with reference to the flowchart of the present invention in FIG.

The signal processing method of the adaptive beamformer of the present invention to be described below is performed by a signal processing system configured to detect a target's azimuth, distance, and depth from a received signal, such as a sonar system.

1) Sensor data FFT step (S1)

This is a step of receiving the target signal of the sound source from each acoustic sensor of the sonar system at a certain direction, and converting the amplitude data of the target signal into frequency data by performing Fast Fourier Transform (FFT).

Next, a covariance matrix between each signal fragment converted into frequency data is calculated, and a covariance matrix is calculated using Equation 1 below.

-K: Covariance matrix

-p (f): Frequency data of signal fragment

-p ^H (f): Hermitian of the frequency data of the signal fragment

The calculation and beamforming of the covariance matrix as described above can be performed using the known homogeneous frequency beamforming technique described above. That is, as illustrated in FIG. 2, a covariance matrix may be calculated by extracting components corresponding to the same frequency ω0 from N signal fragments received by the M-th sensor.

2) Pre-steering stage (S2)

Next, in order to perform steering covariance matrix estimation for each subband, a pre-steering step is performed.

Specifically, the S1 step is performed to pre-steering the converted frequency data at a steering angle pre-assumed for each frequency to calculate the line steering frequency data, and Equation 2 below is used.

-W _M : Time delay vector

-ω: Angular frequency

-θ: steering angle

-T; Transposition matrix

-S: reciprocal of sound speed

-Z: Distance from the first receiver to the receiver

3) Sub-band division step (S3)

The sub-steering frequency data calculated by performing the step S2 is divided for each frequency to calculate a subband matrix, and Equation 3 below is used.

-D: Line steering matrix

-ω: Angular frequency

-θ: steering angle

-W: Time delay term

4) Steering covariance matrix estimation step by sub-station (S4)

The steering covariance matrix for each subband is estimated by using the subband matrix calculated by performing step S3, and Equation 4 below is used.

-K _stmvSub : Steering covariance matrix by sub-station

-ω: Angular frequency

-θ: steering angle

-n: 1,2, ....., N

-N: Number of sub-stations

-B: Frequency frequency of sub-bands

-D ^H : Hermitian of subband matrix

-K: steering covariance matrix

-D: Line steering matrix

5) Adaptive weight vector calculation step for each sub-band (S5)

Step S4 is a step of calculating an adaptive weight vector for each sub-band using the estimated steering covariance matrix for each sub-band, and using Equation 5 below.

-W _stmvSub : Adaptive weight vector by sub-station

-ω: Angular frequency

-θ: steering angle

-n: 1,2, ....., N

-K _stmvSub : Steering covariance matrix by sub-station

-M: 1,2, ....., M

6) Narrow band beam output calculation step (S6)

The step S5 is a step of calculating a narrowband beam output using an adaptive weight vector for each subband calculated by using the step S5, and Equation 6 below is used.

-P: narrowband beam output

-ω: Angular frequency

-θ: steering angle

-W _stmvSub : Adaptive weight vector by sub-station

-D ^H : Hermitian of the line steering matrix

-K: steering covariance matrix

-D: Line steering matrix

7) Broadband beam output calculation step (S7)

It is a step of calculating the broadband beam output by summing all the narrow-band beam outputs calculated by performing step S6.

Therefore, the signal processing method of the adaptive beamformer using the sub-band steering covariance matrix of the present invention, which is composed of the above steps, has a configuration that calculates the narrow-band and wide-band beam output using the sub-band steering covariance matrix. By solving the problem of occurrence of singular matrices due to the elimination of dependencies and the inverse matrix operation, an effect of improving the orientation resolution of the adaptive beamformer was obtained.

The following shows an example of the source code of a computer program that performs the signal processing method of the adaptive beamformer of the present invention as described above.

The source code of Table 1 below is an example of the source code of a computer program that performs calculation of the FFT step and the covariance matrix of the sensor data FFT step (S1).

clearvars; clc;

Src;

%
% (1) fft
%

delf = fs / nfft;
freqX = 0: delf: delf * nfft / 2;

CSDM_vec = zeros (N_snapshot, N_array, nfft / 2);
for i_snapshot = 1: N_snapshot
i_start = overlap * nfft * (i_snapshot-1) +1;
i_end = i_start + nfft-1;
disp (['i_start:' num2str (i_start) '/ i_start:' num2str (i_end) '/' num2str (i_snapshot)])
e_n_snap = e_n (:, i_start: i_end);
for nn = 1: N_array
e_n_snap (nn,:) = e_n_snap (nn,:). * w_t. ';
end

clear n_temp;

e_n_fft = fft (e_n_snap ', nfft) / nfft * 2;
for ll = 1: (nfft / 2)

CSDM_vec (i_snapshot,:, ll) = e_n_fft (ll, :);
end
end

%
% (2) Calculate covariance matrix
%

CSDM = zeros (N_array, N_array, nfft / 2);

for II = 1: nfft / 2
for nn = 1: N_snapshot
CSDM (:,:, II) = CSDM (:,:, II) + CSDM_vec (nn,:, II) '* CSDM_vec (nn,:, II); Equation 1
end
CSDM (:,:, II) = CSDM (:,:, II) / N_snapshot;
end

t_heta = -90: 90;
t_heta = t_heta * pi / 180;
B_m = [cos (t_heta ') sin (t_heta')]; % x, y position

BAND_INTV = (N_array * 3);
CSDM_STMV = zeros (N_array, N_array, length (t_heta), BAND_INTV);
nBandidx = 1;

In addition, Table 2 below is an example of the source code of the computer program that performs the pre-steering step (S2), the sub-band dividing step (S3), and the steering covariance matrix estimation step (S4) for each sub-band.

%
% (3) Pre-steering, sub-band division and steering covariance matrix estimation by sub-band
%


for II = 1: nfft / 2
disp (['' num2str (II) '/' num2str (nfft / 2) ''])
if (II> (nBandidx) * BAND_INTV) && (nBandidx <BAND_INTV);
nBandidx = nBandidx + 1;
end
for m = 1: length (t_heta)
time_delay_steering = -E_n * B_m (m,:) '/ sound_speed;
phase_steering1 = 2 * pi * freqX (II) * time_delay_steering;
w_rep1 = w_a. * exp (-i * phase_steering1); Equation 2
diag_w1 = diag (w_rep1); Equation 3
CSDM_STMV_temp = diag_w1 '* CSDM (:,:, II) * diag_w1; Equation 4
CSDM_STMV (:,:, m, nBandidx) = CSDM_STMV (:,:, m, nBandidx) + CSDM_STMV_temp; Equation 4
end
end

In addition, Table 3 below is an example of the source code of a computer program that performs the adaptive weight vector calculation step (S5), the narrowband beam output calculation step (S6), and the broadband beam output calculation step (S7) for each subband.

%
% (4) Calculation of adaptive weight vector by sub-band and narrow-band beam output
%

time_delay_steering = zeros (N_array, length (t_heta));

for m = 1: length (t_heta)
time_delay_steering (:, m) = -E_n * B_m (m,:) '/ sound_speed; % (N_array * 1)
end

dB_beam_patten1 = zeros (nfft / 2,181,181);
w_stmv1 = zeros (N_array, length (t_heta));
P1 = zeros (nfft / 2, length (t_heta));
norm = zeros (nfft / 2, length (t_heta));

nBandidx = 1;
nBandidx_old = 0;
for II = 1: nfft / 2
disp (['' num2str (II) '/' num2str (nfft / 2) ''])
if (II> (nBandidx) * BAND_INTV) && (nBandidx <BAND_INTV);
nBandidx = nBandidx + 1;
end
for m = 1: length (t_heta)
phase_steering1 = 2 * pi * freqX (II) * time_delay_steering (:, m);
w_rep1 = w_a. * exp (-i * phase_steering1);
diag_w1 = diag (w_rep1);
if nBandidx_old ~ = nBandidx;
w_stmv1 (:, m) = inv (CSDM_STMV (:,:, m, nBandidx)) * ones (N_array, 1) / (ones (N_array, 1) '* inv (CSDM_STMV (:,:, m, nBandidx)) * ones (N_array, 1)); Equation 5
end
norm (II, m) = w_stmv1 (:, m) '* w_stmv1 (:, m);
P1 (II, m) = w_stmv1 (:, m) '* diag_w1' * CSDM (:,:, II) * diag_w1 * w_stmv1 (:, m); Equation 6

end
nBandidx_old = nBandidx;
end

%
% (5) Broadband beam power calculation
%
P2 = sum (P1);

10; Acoustic sensor
11: Signal receiver
12: beam former
13: Detector

Claims (10)

In the signal processing method of the adaptive beamformer performed by a signal processing system configured to detect the orientation, distance, depth of the target from the received signal,
Signal processing method of the adaptive beamformer, characterized in that for performing the steering covariance matrix estimation step (S4) for each subband of the received signal to calculate the narrowband and wideband beam output.
According to claim 1,
The steering covariance matrix estimation step for each sub-band (S4),
Signal processing method of the adaptive beamformer, characterized in that estimated using the following equation (4).

[Equation 4]

-K _stmvSub : Steering covariance matrix by sub-station
-ω: Angular frequency
-θ: steering angle
-n: 1,2, ....., N
-N: Number of sub-stations
-B: Frequency frequency of sub-bands
-D ^H : Hermitian of subband matrix
-K: steering covariance matrix
-D: Line steering matrix
According to claim 2,
Before the steering covariance matrix estimation step (S4) for each sub-band,
A sensor data FFT step (S1) of receiving a target signal of a sound source from a sound sensor from a sound sensor and performing Fourier transform of the amplitude data of the target signal to frequency data;
A pre-steering step (S2) of calculating the pre-steering frequency data by pre-steering the frequency data converted by performing the step S1 at a steering angle pre-assumed for each frequency; And
Signal processing method of the adaptive beamformer, characterized in that the sub-band segmentation step (S3) is performed by dividing the line steering frequency data calculated by performing the step S2 for each frequency to calculate a subband matrix.
The method of claim 3,
The sensor data FFT step (S1),
A signal processing method of an adaptive beamformer, wherein a covariance matrix between each signal fragment converted to frequency data is calculated, and a covariance matrix is calculated using Equation 1 below.

[Equation 1]

-K: Covariance matrix
-p (f): Frequency data of signal fragment
-p ^H (f): Hermitian of the frequency data of the signal fragment
The method of claim 3,
The line steering step (S2),
Signal processing method of the adaptive beamformer, characterized in that calculated by the following equation (2).

[Equation 2]

-W _M : Time delay vector
-ω: Angular frequency
-θ: steering angle
-T; Transposition matrix
-S: reciprocal of sound speed
-Z: Distance from the first receiver to the receiver
The method of claim 3,
The sub-band division step (S3),
Signal processing method of the adaptive beamformer, characterized in that calculated by the following equation (3).

[Equation 3]

-D: Line steering matrix
-ω: Angular frequency
-θ: steering angle
-W: Time delay term
According to claim 2,
After the steering covariance matrix estimation step (S4) for each sub-band,
An adaptive weight vector calculation step for each sub-band using the steering covariance matrix for each sub-band estimated by performing the step S4 (S5),
Narrow-band beam output calculation step (S6) for calculating the narrow-band beam output using the adaptive weight vector for each sub-band calculated by performing the step S5 and
A signal processing method of an adaptive beamformer, characterized in that a broadband beam output calculation step (S7) is performed in which the narrow band beam output calculated by performing the step S6 is summed to calculate the broadband beam output.
The method of claim 7,
The adaptive weight vector calculation step for each sub-band (S5),
Signal processing method of the adaptive beamformer, characterized in that calculated by the following equation (5).

[Equation 5]

-W _stmvSub : Adaptive weight vector by sub-station
-ω: Angular frequency
-θ: steering angle
-n: 1,2, ....., N
-K _stmvSub : Steering covariance matrix by sub-station
-M: 1,2, ....., M
The method of claim 7,
The narrow-band beam output calculation step (S6),
Signal processing method of the adaptive beamformer, characterized in that calculated by the following equation (6).

[Equation 6]

-P: narrowband beam output
-ω: Angular frequency
-θ: steering angle
-W _stmvSub : Adaptive weight vector by sub-station
-D ^H : Hermitian of the line steering matrix
-K: steering covariance matrix
-D: Line steering matrix
10. A computer-readable storage medium containing a program for performing a signal processing method of the adaptive beamformer according to any one of claims 1 to 9.

Images