Multi-band fusion demodulation spectrum calculation method based on multi-feature joint weighting

Abstract

The invention discloses a multi-band fusion demodulation spectrum calculation method based on multi-feature joint weighting. The method comprises the following steps: 1, generating N band-pass filter banks which can completely cover all available bands of a signal and have the same bandwidth and mutually overlapped passbands; 2, processing the time domain signal into a band-pass signal xn (t) by using a filter bank, and calculating a demodulation spectrum Dn (f) of the xn (t) by using square detection; 3, calculating skewness Sn and kurtosis Kn of the demodulation spectrum Dn (f) of each frequency band, and determining a fusion weighting coefficient Wn of each frequency band according to the skewness Sn and kurtosis Kn; and step 4, calculating an adaptive fusion demodulation spectrum D (f) according to the weighting coefficient. The method can effectively adapt to the optimal demodulation frequency band difference caused by the marine environment and the target type, greatly improves the demodulation spectrum detection and identification capability of the suspicious target under the condition of low signal-to-noise ratio, and is good in adaptability, simple in processing method and high in reliability.

Description

Technical Field

[0001] The present invention relates to a sonar signal processing method, and belongs to the fields of underwater acoustic engineering, sonar technology, and passive sonar signal processing, and in particular to a multi-band fusion demodulation spectrum calculation method based on multi-feature joint weighting. Background Art

[0002] Underwater targets such as ships have multiple noise sources, such as thrusters, propellers, and various pumps. These noises interact with the surrounding water and propagate through the channel to be detected by sonar. Passive target radiated noise primarily includes mechanical noise, propeller noise, and hydrodynamic noise. Passive identification identifies underwater targets by analyzing and extracting the characteristics of passive target radiated noise. Demodulated spectral lines are a commonly used characteristic measure of underwater target radiated noise. Long-range detection and identification of underwater targets based on target demodulation spectral features has also achieved good results.

[0003] However, in recent years, with the development of advanced noise reduction technologies, the intensity of characteristic line spectra in the demodulation spectrum of underwater targets has decreased. Furthermore, as sonar detection range has increased significantly, the signal-to-noise ratio (SNR) of target radiated noise received by sonar has been significantly reduced. Furthermore, different types of targets often have different optimal demodulation frequency bands, and the appropriateness of the selected demodulation frequency band has a significant impact on the detection of demodulation line spectrum features. In the complex and changing ocean acoustic environment, adaptively selecting demodulation frequency bands based on the acoustic characteristics of the target and calculating the demodulation spectrum is of great significance.

[0004] The classic demodulation spectrum calculation method determines the demodulation frequency band based on prior knowledge. This method works well when dealing with common targets with high signal-to-noise ratios, but its performance degrades severely when dealing with unknown targets with low signal-to-noise ratios due to improper selection of the demodulation frequency band. Summary of the Invention

[0005] Based on the deficiencies of the existing technology, the present invention provides a multi-band fusion demodulation spectrum calculation method based on multi-feature joint weighting to achieve the advantages of strong detection capability, good adaptability, simple processing method and high reliability.

[0006] The technical solution of the present invention is:

[0007] A multi-band fusion demodulation spectrum calculation method based on multi-feature joint weighting, comprising:

[0008] Step 1: Generate N bandpass filter banks with the same bandwidth and overlapping passbands that can fully cover all available frequency bands of the signal;

[0009] Step 2: Use a filter bank to process the time domain signal into a bandpass signal x n (t), calculate x using square detection n The demodulation spectrum D of (t) n (f);

[0010] Step 3: Calculate the demodulation spectrum D of each frequency band n Skewness S of (f) n and kurtosis K n , and determine the fusion weighting coefficient W of each frequency band based on the two n ;

[0011] Step 4: Calculate the adaptive fusion demodulation spectrum D(f) according to the weighting coefficient.

[0012] The advantages of the present invention are: it provides a signal processing method that uses skewness and kurtosis to quantitatively evaluate the quality of demodulated spectra, thereby achieving adaptive and optimal fusion optimization of demodulated spectra. The processing method of the present invention has the characteristics of adaptive multi-band fusion, which is specifically manifested in:

[0013] (1) The present invention utilizes multiple continuous demodulation frequency bands to completely cover all effective frequency bands, ensuring that the target information is fully utilized.

[0014] (2) The present invention introduces the joint statistic of skewness-kurtosis to realize the quantitative evaluation of the demodulation spectrum quality of each frequency band signal. Based on this, a calculation method for adaptively and preferentially fusing the demodulation spectra of each frequency band is proposed to realize the fusion of multiple high-quality frequency band demodulation spectra and generate a fused demodulation spectrum that is better than the demodulation spectrum of a single demodulation frequency band. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a workflow diagram of the present invention;

[0016] Figure 2 is the time domain waveform of the signal to be processed;

[0017] Figure 3 is the demodulated spectrum of the signal frequency band 1 to be processed;

[0018] Figure 4 is the demodulated spectrum of the signal frequency band 2 to be processed;

[0019] Figure 5 is the demodulated spectrum of the signal frequency band 3 to be processed;

[0020] Figure 6 is the demodulated spectrum of the signal frequency band 4 to be processed;

[0021] Figure 7 is the fused demodulated spectrum of the signal to be processed. DETAILED DESCRIPTION

[0022] The present invention will be further described below with reference to specific embodiments and accompanying drawings:

[0023] The present invention provides a multi-band fusion demodulation spectrum calculation method based on multi-feature joint weighting, such as Figure 1The specific steps are as follows:

[0024] Step 1: Generate filter bank:

[0025] For time domain waveforms such as Figure 2 The sample signal x(t) shown in the figure is processed. According to the available frequency band of x(t), N filter banks with bandwidth B and passband overlap rate of 50% are designed. The nth filter is denoted as BPF. n .

[0026] Step 2: Calculate the demodulation spectrum of each frequency band through bandpass filtering and square detection:

[0027] Bandpass filter the sample signal x(t) to obtain the frequency band signal x n (t) are as follows:

[0028] x n (t) = BPF n {x(t)} (1)

[0029] Among them, BPF n {·} represents the use of the nth bandpass filter to process the signal.

[0030] x n (t) Perform square detection to obtain the detection signal v n (t) are as follows:

[0031] v n (t) = [x n (t)] 2 (2)

[0032] v n (t) contains x n (t) square frequency component and modulation frequency component. Low-pass filtering is performed to remove the high-frequency square frequency component and retain the required modulation frequency component. The cutoff frequency and parameter design of the low-pass filter are determined based on engineering practice and prior knowledge, and then the spectrum is analyzed using fast Fourier transform. The above process obtains the demodulated spectrum D n The mathematical expression of (f) is

[0033] D n (f) = FFT {LPF {v n (t)}} (3)

[0034] Among them, FFT{·} represents fast Fourier transform, LPF{·} represents low-pass filtering, and the demodulation spectrum of each frequency band is as follows Figure 3 、 Figure 4 、 Figure 5 、 Figure 6 shown.

[0035] In practical engineering, one can first analyze the characteristics of the signal to be processed to determine the frequency range of the useful modulation frequency component. Alternatively, one can draw on past experience or industry standards to determine a universal frequency range, the upper bound of which serves as the low-pass filter's cutoff frequency. Depending on the application requirements, Butterworth filter parameters may be used when excessive smoothing of the passband and stopband is required, Chebyshev filter parameters may be used when excessively steepening of the passband and stopband is required, or elliptic filters may be used when computing resources are limited. Based on these criteria, a low-pass filter is designed and the signal is filtered.

[0036] Step 3: Calculate the skewness and kurtosis of the demodulated spectrum of each frequency band and generate the fusion weighting coefficient:

[0037] Calculate the demodulation spectrum D n Spectral skewness of (f):

[0038]

[0039] And the spectral kurtosis:

[0040]

[0041] Among them, D n (f i ) represents the frequency band of interest in the demodulated spectrum, represents the mean of the frequency band of interest, s n represents the standard deviation of the frequency band of interest, and I represents the total number of frequency points in the frequency band of interest.

[0042] The calculated skewness and kurtosis are regularized to ensure the effectiveness of weighted fusion. The formula is as follows:

[0043]

[0044]

[0045] Based on the regularized skewness and kurtosis, the joint weighted coefficient is calculated as follows:

[0046]

[0047] Step 4: Weighted fusion generates adaptive optimal demodulation spectrum:

[0048] Calculate the skewness-kurtosis joint weighted adaptive fusion demodulation spectrum:

[0049]

[0050] At this time, the spectrum of the adaptive fusion demodulation spectrum D(f) is as follows: Figure 7 shown.

[0051] Observation and comparison Figures 3 to 6The classical demodulation spectrum of multiple frequency bands and Figure 7 The multi-band fusion demodulation spectrum based on multi-feature joint weighting. It can be found that Figures 3 to 6 The harmonic structure of the Figure 7 Harmonics are still observable above 30 Hz, and a harmonic with a difference frequency of 2.7 Hz can be directly observed across the entire frequency band. This phenomenon demonstrates that the proposed method utilizes the skewness-kurtosis of multi-band demodulation spectra to achieve adaptive weighted fusion, resulting in a demodulation spectrum that outperforms classical demodulation methods, enhances characteristic line spectra, and facilitates further analysis and signal identification.

[0052] In summary, the present invention proposes a multi-band fusion demodulation spectrum calculation method based on multi-feature joint weighting, which first calculates the demodulation spectrum of multiple continuous frequency bands of the signal, and then uses skewness and kurtosis to quantitatively evaluate and preferentially fuse the demodulation spectrum of each frequency band, thereby realizing adaptive selection of the optimal demodulation frequency band, and then realizing demodulation spectrum enhancement under low signal-to-noise ratio conditions. The demodulation spectrum calculation method provided by the present invention has full-band coverage and optimal frequency band adaptability, and realizes full-band coverage by calculating the demodulation spectrum of multiple continuous frequency bands, and realizes preferential fusion of multi-band demodulation spectrum by adaptively selecting high-quality demodulation frequency bands based on the joint features of skewness and kurtosis. The method of the present invention can effectively adapt to the differences in optimal demodulation frequency bands caused by the marine environment and target type, greatly improves the demodulation spectrum detection and identification capabilities of suspicious targets under low signal-to-noise ratio conditions, has good adaptability, simple processing method, and high reliability.

[0053] It should be noted that the above embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Equivalent modifications made based on the above embodiments fall within the scope of protection of the present invention.

Claims

1. A multi-band fusion demodulation spectrum calculation method based on multi-feature joint weighting, characterized in that: include: Step 1: Generate N bandpass filter banks with the same bandwidth and overlapping passbands that can fully cover all available frequency bands of the signal; Step 2: Use a filter bank to process the time domain signal into a bandpass signal x n (t), calculate x using square detection n The demodulation spectrum D of (t) n (f); Step 3: Calculate the demodulation spectrum D of each frequency band n Skewness S of (f) n and kurtosis K n , and determine the fusion weighting coefficient W of each frequency band based on the two n ; Step 4: Calculate the adaptive fusion demodulation spectrum D(f) according to the weighting coefficient.

2. The method for calculating a multi-band fusion demodulation spectrum based on multi-feature joint weighting according to claim 1, characterized in that: In the step 1, the time domain waveform signal x(t) is processed, and N filter banks with a bandwidth of B and a passband overlap rate of 50% are designed according to the available frequency band of x(t), where the nth filter is denoted as BPF. n .

3. The method for calculating a multi-band fusion demodulation spectrum based on multi-feature joint weighting according to claim 1 or 2, characterized in that: In the second step, the sample signal x(t) is band-pass filtered to obtain the frequency band signal x n (t) are as follows: x n (t)=BPF n {x(t)} (1) Among them, BPF n {·} represents the use of the nth bandpass filter to process the signal; x n (t) Perform square detection to obtain the detection signal v n (t) are as follows: v n (t)=[x n (t)] 2 (2) v n (t) contains x n The square frequency component and modulation frequency component of (t); Then perform low-pass filtering to remove the high-frequency square frequency component and retain the required modulation frequency component, determine the cutoff frequency and parameter design of the low-pass filter, and then use fast Fourier transform to perform spectrum analysis on the low-pass filtered signal. The above process obtains the demodulated spectrum D n The mathematical expression of (f) is: D n (f)=FFT{LPF{v n (t)}} (3) Wherein, FFT{·} represents fast Fourier transform, and LPF{·} represents low-pass filtering.

4. The method for calculating a multi-band fusion demodulation spectrum based on multi-feature joint weighting according to claim 3, characterized in that: In step 3, the demodulation spectrum D is calculated. n Spectral skewness of (f): And the spectral kurtosis: Among them, D n (f i ) represents the frequency band of interest in the demodulated spectrum, represents the mean of the frequency band of interest, s n represents the standard deviation of the frequency band of interest, and I represents the total number of frequency points in the frequency band of interest; The calculated skewness and kurtosis are regularized to ensure the effectiveness of weighted fusion. The formula is as follows: Based on the regularized skewness and kurtosis, the joint weighted coefficient is calculated as follows:

5. The method for calculating multi-band fusion demodulation spectrum based on multi-feature joint weighting according to claim 4, characterized in that: In the step 4, the skewness-kurtosis joint weighted adaptive fusion demodulation spectrum is calculated: The frequency spectrum of the adaptive fused demodulated spectrum D(f) is obtained.

Images