A line spectrum detection method utilizing the polarization characteristic parameters of a single vector sensor

Abstract

This invention relates to a line spectrum detection technique for ship targets under strong background noise interference in the field of underwater acoustic signal processing. It is applicable to the detection of line spectrum signals when they are severely affected by environmental noise or the Puller effect. The underwater target line spectrum signal received by a single-vector sensor exhibits polarization characteristics, while the polarization characteristics of random noise differ from those of the signal. Research on this polarization characteristic in the field of underwater acoustic signal processing is limited. Therefore, this invention calculates the spectral matrix of the line spectrum signal received by the single-vector sensor, performs singular value decomposition on it, and proposes suitable polarization characteristic parameters for line spectrum signal detection. This invention can significantly reduce interference from Gaussian white noise, Gaussian colored noise, and actual marine environmental noise, providing a new approach for the detection of weak line spectrum signals. Furthermore, this invention is simple to operate, has a moderate computational load, and produces clear and stable results, making it widely applicable.

Description

Technical Field

[0001] This invention relates to the field of underwater acoustic signal processing, and in particular to the field of ship radiated noise signal detection. Background Technology

[0002] The line spectrum of radiated noise from ships is typically 10-25 dB higher than the average spectral level of the continuous spectrum, and is often used as a detection signal for detecting ship targets. The commonly used line spectrum detection technique for ship targets is classical spectrum estimation, which acquires line spectrum features by accumulating the spectrum of the line spectrum signal over a long period. However, underwater ship targets are affected by marine environmental noise, which introduces severe background noise interference to the line spectrum detection of ship targets. Simultaneously, the Doppler effect generated by ships during navigation makes the classical spectrum estimation technique less effective for accumulating the line spectrum of ship targets. Therefore, how to detect the line spectrum signal of underwater ship targets in the complex marine environment has always been a pressing problem in the field of line spectrum detection.

[0003] To detect line spectrum signals of ship targets in complex marine environments, researchers have proposed a series of line spectrum detection techniques, such as adaptive line spectrum enhancement, higher-order cumulant techniques, and frequency-domain phase standard deviation weighting techniques. These techniques utilize the amplitude, phase, and frequency information of the target signal for line spectrum detection, which can suppress interference from marine environmental noise to a certain extent and still detect line spectrum signals. Using a single-vector sensor to receive signals from underwater ship targets not only improves the sensor's resistance to isotropic noise interference but also allows for the acquisition of vector information of the ship target, and the reception method is simple. However, vector signals have polarization characteristics, and the polarization characteristics of random noise often differ from those of the signal itself; however, research on this characteristic in the field of underwater acoustic signal processing is limited. Therefore, it is necessary to study the appropriate polarization characteristic parameters for detecting line spectrum signals received by single-vector sensors. Summary of the Invention

[0004] The purpose of this invention is to develop a line spectrum detection technology for ship targets under strong marine background noise interference, aiming to solve the problems of line spectrum signals being severely affected by marine environmental noise and difficult to detect due to the Doppler effect. Existing underwater target line spectrum signal detection technologies can detect line spectrum signals under certain signal-to-noise ratios, but they are highly dependent on the estimation error of strong marine background interference and have poor suppression capabilities. This invention utilizes the polarization characteristics of the vector signal received by a single vector sensor to stably and quickly detect line spectrum signals from strong marine background interference. Compared with other line spectrum detection technologies for underwater ship targets, this invention has a weaker dependence on the estimation error of strong marine background noise interference and can effectively suppress strong marine environmental background interference, thereby detecting line spectrum signals.

[0005] The principles upon which this invention is based are as follows:

[0006] Due to the relative motion between the ship target and the receiving sensor, or the instability of the seawater channel, the Doppler effect occurs, and the line spectrum signal of the ship target in the water exhibits dynamic characteristics. Assuming a far-field scenario, the line spectrum signal is incident on a single-vector sensor with x and y channels, and the single-vector sensor is placed parallel to the ground, the received line spectrum signal data model is as follows:

[0007]

[0008] have:

[0009]

[0010] In the formula, θ is the azimuth angle of the incident line spectrum signal; P is the total number of sinusoidal signals of different frequencies; For random initial phase; A i f is the amplitude of the sinusoidal signal; i Here, is the frequency of the line spectrum signal, measured in Hz, and i is the index of the line spectrum signal frequency; n w,x (t) and n w,y (t) represents the Gaussian white noise received by the x and y channels, respectively, and n c,x (t) and n c,y (t) represents the Gaussian colored noise received by the x and y channels, respectively. Δf i The Doppler frequency shift of the line spectrum signal is related to the line spectrum signal as follows:

[0011]

[0012] In the formula, v represents the relative velocity between the ship target and the single vector sensor, in m / s; c is the speed of sound in water.

[0013] set up

[0014]

[0015]

[0016] The vector signal received by the vector sensor can then be written in the following form:

[0017] r(t)=s(t)+n(t) (7)

[0018] Performing a Fourier transform on the signal r(t) received by the single-vector sensor yields the following spectrum matrix along the positive frequency axis:

[0019] R(f) = S(f) + N(f)

[0020]

[0021] In the formula, R(f) is the spectrum matrix obtained after Fourier transform, S(f) is the spectrum matrix corresponding to the line spectrum signal, N(f) is the spectrum matrix corresponding to the marine background noise interference, and B i,x B i,y These represent the amplitude of the line spectrum signal received in each channel.

[0022] With frequency f as the center and bandwidth Δ as the bandwidth, R(f) is divided into frequency bands, and then the spectral matrix R within the frequency band is processed. Δ (f) Perform singular value decomposition. The singular value decomposition formula is as follows:

[0023]

[0024] In the formula, T represents the transpose operation, and R Δ (f) is a 2×N matrix, where N is the number of frequency units within the bandwidth Δ; U is an N×N unitary matrix, and vector u m (m=1、2) Corresponding to The characteristic column vectors; D is an N×2 matrix with singular values ​​d j =D j,j (j=1、2) represent the magnitudes of the major and minor axes of the spectral amplitude distribution graph at each frequency point within the corresponding frequency band; V is a 2×2 unitary matrix, and its vector v m correspond eigenvectors.

[0025] Assuming the line spectrum signal data received by the single-vector sensor is infinitely long, and neglecting noise within the frequency band containing the line spectrum signal, we have: Since d2 = 0, the energy of the line spectrum signal is reflected in the magnitude of d1. Because the line spectrum signal actually received by a single-vector sensor is of finite length, the data covariance between channels within the frequency band is not zero. Therefore, the energy of the line spectrum signal always exists within the frequency band containing d1. s,1 >d s,2 The relationship ≠ 0 exists, that is, d s,1 -d s,2 >0, where d s,1 and d s,2 This represents the singular values ​​within the frequency band of the line spectrum signal; however, within the frequency band containing only noise, because the correlation between Gaussian white noise or Gaussian colored noise channels is smaller than that of the line spectrum signal, the spectral amplitude distribution of Gaussian white noise at each frequency point should approximate a circle, while the spectral amplitude distribution of Gaussian colored noise at each frequency point is an ellipse, with d s,1 -d s,2 >>d n,1 -d n,2 , where d n,1 and d n,2 This represents the singular values ​​within the frequency band of the noise spectrum. Therefore, the proposed polarization characteristic parameters are:

[0026] η f =(d1-d2) 2 (10)

[0027] As can be seen from equation (10), under a certain signal-to-noise ratio, the present invention has a relatively small relationship with the intensity of marine background noise, but a strong relationship with the magnitude of singular values ​​in the frequency band.

[0028] The specific operation method of this invention is as follows:

[0029] (1) Perform a Fourier transform on the line spectrum signals received by each channel of the single vector sensor to obtain the spectrum matrix, i.e., calculate R(f), where f is the frequency. Essentially, this involves calculating the discrete Fourier transform of each channel of the single vector sensor, transforming the time-domain signal into a frequency-domain signal;

[0030] (2) With frequency f as the center and Δ as the bandwidth, perform frequency banding on R(f);

[0031] (3) For the spectrum matrix R within the frequency band Δ (f) Perform singular value decomposition to obtain the maximum singular value d1 and the minimum singular value d2 in each frequency band;

[0032] (4) Calculate the polarization characteristic parameter η(f) in each frequency band according to equation (10);

[0033] (5) By performing cubic spline interpolation on η(f), the polarization characteristic parameters at each frequency point can be obtained, and the proposed polarization characteristic parameters can be used to detect the line spectrum signal.

[0034] The present invention has the following advantages:

[0035] 1. This invention can stably and rapidly detect line spectrum signals received by a single vector sensor. Furthermore, compared to classical power spectrum estimation techniques, the detection probability of polarization characteristic parameters is significantly better than that of classical spectrum estimation techniques. Therefore, this invention is of great significance for the detection of weak line spectrum signals.

[0036] 2. This invention can greatly reduce the interference of Gaussian white noise, Gaussian colored noise and actual marine environmental noise. Therefore, this invention is of great significance for noise suppression technology.

[0037] 3. The present invention is simple to operate, has a moderate amount of computation, and is widely applicable. It can play a role in the detection of weak line spectrum signals under strong background noise interference and has high application value. Attached Figure Description

[0038] Figure 1 Invention content flowchart;

[0039] Figure 2 x-channel line spectrum signal spectrum;

[0040] Figure 3 y-channel line spectrum signal spectrum;

[0041] Figure 4 Maximum singular value within each frequency band of the line spectrum signal;

[0042] Figure 5 Minimum singular value within each frequency band of a line spectrum signal;

[0043] Figure 6 Detection results of polarization characteristic parameters;

[0044] Figure 7 Detection probability results of polarization characteristic parameter line spectrum (solid line); Detection probability results of power spectrum line spectrum of two channels (dashed line);

[0045] Figure 8 Experimental data processing results. Detailed Implementation

[0046] The invention will now be described in more detail with reference to the accompanying drawings:

[0047] (1) The operation flow of the present invention using programming to implement functions is as follows: Figure 1 As shown. First, a Fourier transform is performed on the data received by each channel of the single-vector sensor to obtain the spectrum matrix of the data. Essentially, this involves calculating the discrete Fourier transform of each channel of the single-vector sensor, converting the time-domain signal into a frequency-domain signal. Assume the length of the ship target line spectrum signal received by each channel of the single-vector sensor is 10s, the sampling frequency is 2000Hz, the incident azimuth angle varies from 90° to 10° in steps of 5°, the frequencies of the line spectrum signals are f1 = 420Hz, f2 = 470Hz, f3 = 570Hz, the relative velocity between the ship target and the single-vector sensor is 15.4m / s, and the power of the signal and the ratio of the power of Gaussian white noise to Gaussian colored noise are 5dB and 10dB, respectively. The frequency band of the Gaussian colored noise is between 430Hz and 435Hz. A Fourier transform is performed on the line spectrum signal of each channel to obtain the spectrum of the line spectrum signal of each channel, as shown below. Figure 2 and Figure 3 As shown. Figure 2 and Figure 3 The results show that the signal-to-noise ratio of the three line spectrum signals in the x channel is low, and there is obvious Gaussian colored noise interference between 430Hz and 435Hz; while the signal-to-noise ratio of the three line spectrum signals in the y channel is high, and there is obvious Gaussian colored noise interference between 430Hz and 435Hz, and it is difficult to distinguish between the line spectrum signals and the Gaussian colored noise.

[0048] (2) Construct a spectrum matrix from the spectrum of the single-vector sensor obtained in (1), then perform frequency banding on the spectrum matrix, and perform singular value decomposition on the spectrum matrix within each frequency band to obtain the maximum and minimum singular values ​​within each frequency band, such as Figure 4 and Figure 5 As shown. Figure 4 The results show that the maximum singular value in the frequency band where the line spectrum signal is located and the maximum singular value in the frequency band where the Gaussian colored noise is located are much larger than the maximum singular value in other frequency bands. Figure 5 The results show that the minimum singular value in the frequency band containing Gaussian colored noise is much larger than the minimum singular value in other frequency bands. The polarization characteristic parameters in each frequency band are calculated according to equation (10), and cubic spline interpolation is performed on the polarization characteristic parameters to obtain the polarization characteristic parameters at each frequency point, such as... Figure 6 As shown. Figure 6 The results show that the polarization characteristic parameter values ​​within the frequency band of the line spectrum signal are much larger than those within the frequency bands of Gaussian white noise and Gaussian colored noise, indicating that the line spectrum signal can be detected effectively using these polarization characteristic parameters. This is consistent with the previous theoretical analysis results.

[0049] (3) Figure 7 The results show a comparison of the detection probabilities of polarization characteristic parameters and power spectra obtained using classical spectral estimation techniques at different signal-to-noise ratios. Figure 7 The results show that the detection results of polarization characteristic parameters are significantly better than those of the power spectrum method for each channel. When the signal-to-noise ratio is -2dB, the detection probability of polarization characteristic parameters can reach 100%; when the signal-to-noise ratio is -5dB, the detection probability of polarization characteristic parameters can reach 94%, while the detection probability of power spectrum is only 32%. Figure 8 This is the result of processing data from a sea trial experiment. The frequency of the line spectrum signal is known to be 275Hz. Figure 8 (a) and Figure 8 (b) The power spectrum line spectrum detection results of the x and y channel line spectrum signals are respectively. Figure 8 (c) shows the detection results of polarization characteristic parameters. Figure 8 The results show that the signal-to-noise ratio (SNR) of the 275Hz line spectrum signal in the x-channel is low, and colored noise interference exists around 260Hz and 300Hz, making it impossible to detect the line spectrum signal. The SNR of the 275Hz line spectrum signal in the y-channel is higher than that in the x-channel, but the line spectrum detection performance is poor. Figure 8 In (c), the polarization characteristic parameter value corresponding to the actual marine environmental noise is significantly smaller than the polarization characteristic parameter value at the line spectrum signal, and the line spectrum signal with a frequency of 275Hz can be easily detected.

Claims

1. A line spectrum detection method using a polarization signature parameter of a single vector sensor, characterized in that, The following steps are taken: 1) Fourier transform the line spectrum signal received by each channel of the single vector sensor to obtain a frequency spectrum matrix, and calculate R(f), wherein f is the frequency; that is, perform discrete Fourier transform on the data of each channel of the single vector sensor to convert the time domain signal to a frequency domain signal; 2) Perform frequency band processing on R(f) with frequency f as the center and Δ as the bandwidth; 3) a spectral matrix R within a frequency band Δ (f) performing singular value decomposition to obtain a largest singular value di and a smallest singular value d2 within each frequency band; 4) Calculate the polarization signature η within each frequency band according to equation (10) f ; 5) for η f The polarization characteristic parameters at each frequency point can be obtained by performing three times of spline interpolation, and the linear spectrum signal is detected by using the proposed polarization characteristic parameters. The polarization characteristic parameter η in step 4) f is obtained by the following process: Considering the influence of Doppler effect and strong background noise on the signal received by the single vector sensor; assuming that the line spectrum signal is incident on the single vector sensor with x and y channels in the far field, the single vector sensor is placed parallel to the ground, and the data model of the received line spectrum signal is There are In the formula, θ is the azimuth angle of the incident line spectrum signal; P is the total number of sinusoidal signals at different frequencies; is a random initial phase; A i is the amplitude of the sinusoidal signal; f i f is the frequency of the line spectrum signal in Hz; i is the serial number of the line spectrum signal frequency; n w,x (t) and n w,y (t) represent the Gaussian white noise received by the x, y two channels, respectively; n c,x (t) and n c,y (t) represent the Gaussian color noise received by x, y channels, respectively; Δf i Doppler shift for a line spectrum signal, which has the following relationship with the line spectrum signal: In the formula, v represents the relative motion speed of the ship target and the single vector sensor, and the unit is m / s; c is the propagation speed of sound in water; Let The vector signal received by the vector sensor can be written as follows: r(t) = s(t) + n(t) (7) Fourier transform the signal r(t) received by the single vector sensor to obtain the frequency spectrum matrix on the positive frequency axis: In the formula, R(f) is a frequency spectrum matrix obtained after Fourier transform, S(f) is a frequency spectrum matrix corresponding to the line spectrum signal, N(f) is a frequency spectrum matrix corresponding to the ocean background noise interference, B i,x , B i,y respectively represent the amplitude of the line spectrum signal received in each channel. The R(f) is processed in frequency bands with f as the center and Δ as the bandwidth, and the spectral matrix R Δ (f) is singular value decomposed; the singular value decomposition formula is as follows: where T denotes a transposition operation, R Δ (f) is a 2xN matrix, N is the number of frequency units in the bandwidth Δ; U is an N x N unitary matrix, the vector u m (m = 1, 2) corresponds to the characteristic column vector of ; D is an N x 2 matrix, the singular value d j = D j,j (j = 1, 2) corresponds to the length and the short axis of the spectral amplitude distribution pattern at each frequency point in the frequency band; V is a 2 x 2 unitary matrix, the vector v m corresponds to the characteristic vector of The proposed polarization characteristic parameter is: η f = (d1 - d2) 2 (10).

2. The line spectrum detection method of a polarization characteristic parameter of a single vector sensor according to claim 1, characterized in that, When the noise is not considered in the frequency band where the line spectrum signal is located, for the infinite long line spectrum signal data, there are d2=0, it is known that the energy of the line spectrum signal is reflected in the size of d1; since the line spectrum signal actually received by the single vector sensor is of finite length, the data covariance between each channel in the frequency band is not 0, so there is always a relationship of d s,1 >d s,2 ≠0 in the frequency band where the line spectrum signal is located, that is, d s,1 -d s,2 >0, where d s,1 and d s,2 represent singular values in the frequency band where the line spectrum signal is located; In the frequency band where only noise exists, due to the correlation of the Gaussian white noise or the Gaussian color noise between channels being smaller than the correlation of the line spectrum signal, the spectrum amplitude distribution pattern of each frequency point of the Gaussian white noise should be approximately a circle, and the spectrum amplitude distribution pattern of each frequency point of the Gaussian color noise is an ellipse, with d s,1 -d s,2 >>d n,1 -d n,2 wherein d n,1 and d n,2 represent singular values in the frequency band where the noise spectrum exists. Therefore, by using η f = (d1-d2) 2 The line spectrum signal is detected by using both the energy information and the polarization information of the line spectrum signal, so that the line spectrum signal can be detected in the presence of Doppler effect and strong background noise interference.

Images