A method for extracting internal wave acoustic features based on energy jitter

Abstract

The present invention discloses a method for extracting internal wave acoustic features based on energy jitter, comprising the following steps: S100, preprocessing of acoustic signals in the internal wave region: filtering and noise cancellation processing are sequentially performed on the multi-channel acoustic signals acquired by the hydrophone, and then a time-frequency diagram of the acoustic signal is obtained through a windowed short-time Fourier transform; S200, extraction of internal wave acoustic energy jitter features: acquiring data for a specific time period based on the acoustic signal time-frequency diagram, processing and analyzing the frequency point energy, and obtaining an energy jitter spectrum. The present invention has high accuracy: by comprehensively preprocessing the multi-channel acoustic signals, low-frequency and high-frequency signals as well as vehicle noise and background noise are effectively suppressed, making the subsequently extracted internal wave acoustic features more accurate and reliable. The present invention has strong timeliness: the present invention can more efficiently extract internal wave acoustic energy jitter features from the time-frequency image, thereby improving the speed of internal wave information processing to a certain extent.

Description

Technical Field

[0001] The present invention relates to the technical field of ocean information acquisition, and in particular to a method for extracting internal wave acoustic features based on energy jitter. Background Art

[0002] In the study of ocean internal waves, the theory of the interaction between surface waves and internal waves, as well as the acquisition of detectable characteristics of the resulting internal waves, are among the most important frontiers in ocean science research. The interaction between surface waves and internal waves involves complex theories of wave-current interaction and wave-wave interaction, which are directly related to the near-surface flow of wind fields and, therefore, are also a key area of ​​ocean dynamics research.

[0003] The interaction between surface waves and internal waves affects the local characteristics of the wave spectrum. Currently, analysis of these spectral characteristics cannot be fully addressed using methods such as fully developed wave field spectra. Modification of the source term of the wave spectrum may even be required, impacting the quantification of remote sensing in the internal wave region. In ocean remote sensing research, the commonly used method for calculating wave breaking caused by the interaction between surface waves and internal waves is to calculate the wave foam coverage. However, this method also presents some technical difficulties that are difficult to overcome due to the uneven distribution of wave foam in the internal wave region.

[0004] Furthermore, difficulties remain in the research on the characteristic detection of internal waves, especially in the detection of internal waves on underwater platforms. Due to technical constraints such as real-time performance and efficient underwater communications, internal wave observation can be achieved through various means, such as adding CTDs at appropriate locations and densities to submerged buoys for on-site observation of internal waves, using ADCPs to observe flow fields in internal wave zones, and using optical or microwave (SAR) high-resolution satellites to observe ocean internal waves over a wide area. While these methods have solved the problem of internal wave monitoring to a certain extent, due to the limitations of traditional underwater communication technology, it is still difficult for underwater platforms to obtain internal wave information in real time. Existing observation methods cannot fully and effectively achieve high-time efficiency in processing internal wave information. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for extracting internal wave acoustic features based on energy jitter, which is used for real-time detection and early warning of ocean internal waves by underwater working platforms, and solves the problems of existing technologies in internal wave acoustic feature extraction and information processing timeliness.

[0006] The present invention is implemented as follows: a method for extracting internal wave acoustic characteristics based on energy jitter, such as Figure 10 As shown, the following steps are included:

[0007] S100, preprocessing of acoustic signals in the internal wave region: performing filtering and noise cancellation processing on the multi-channel acoustic signals acquired by the hydrophone in sequence, and then obtaining a time-frequency diagram of the acoustic signals through a windowed short-time Fourier transform;

[0008] S200 , extracting internal wave acoustic energy jitter characteristics: acquiring data for a specific time period based on the acoustic signal time-frequency diagram, processing and analyzing the frequency point energy, and obtaining an energy jitter spectrum.

[0009] Furthermore, in step S100, the filtering process is to use an FIR bandpass filter designed based on a Kaiser window to suppress low-frequency and high-frequency signals of the multi-channel acoustic signal.

[0010] Furthermore, the Kaiser window function is shown as follows: in, is the window length. It is an adjustable parameter whose shape can be controlled and can directly affect the trade-off between sidelobe attenuation and mainlobe width. It is a zero-order first-kind modified Bessel function and can reflect the smooth transition characteristics of the window function. The empirical formula is as follows The window length can be estimated using the following formula: in, is the normalized transition bandwidth, is the stop-band attenuation;

[0011] The steps to design an FIR bandpass filter based on the Kaiser window include:

[0012] S111. Given filter indicators, including stopband frequency, passband frequency, stopband attenuation, and ripple parameters;

[0013] S112, calculation :according to Using the above empirical formula;

[0014] S113, confirm : Use the transition bandwidth formula to calculate the required window length;

[0015] S114, generate window function: substitute Kaiser window formula to generate time domain window;

[0016] S115. Multiply the window function by the ideal filter impulse response.

[0017] Furthermore, in step 100, the noise cancellation processing is to use an LMS adaptive filter to perform two cancellation filters. The first time, the vehicle self-noise channel is used as the reference channel, and the forward and side channel data are adaptively processed to suppress the vehicle noise. The second time, the processed side channel signal is used as the reference channel to filter the forward channel to cancel the background noise.

[0018] Furthermore, the LMS cancellation processing algorithm of the internal wave signal model is as follows: in , respectively represent the forward and side channels of the hydrophone on the vehicle; The target signal output from the side channel is recorded as the reference signal, and the target signal output from the forward channel is recorded as the new signal to be processed, and the LMS cancellation processing algorithm is applied again; here, is the self-noise measurement value of the vehicle, is the output signal of the filter, is the forward channel signal, is the target signal after denoising the forward channel signal; is the weight vector, is the step size parameter, To adjust the parameters;

[0019] The process of filtering by the LMS adaptive filter includes the following steps:

[0020] S121. Initialize filter parameters: Before starting adaptive filtering, initialize the filter parameters and set the initial value of the filter to 0.

[0021] S122. Read the collected multi-channel signals: including noise signals , forward and side channel signals and the signal of channel 2 .

[0022] S123 , calculating the estimated signal and the error signal by using a weighted sum of filter coefficients, and adjusting and updating the filter parameters accordingly.

[0023] S124. Repeat the above steps and continuously update the filter parameters through multiple iterations until the error of the filter output signal meets the preset requirements.

[0024] Furthermore, the mathematical expression of short-time Fourier transform is: in is the original signal, is a window function; the windowed short-time Fourier transform step comprises:

[0025] S131, framing and windowing processing: Divide the forward channel filtered signal into pieces of length The short time frame of ; In order to reduce spectrum leakage, Hamming window processing is further adopted here; the formula is ;

[0026] S132, Short-time Fourier transform: Perform fast Fourier transform (FFT) on each frame of signal to obtain the complex spectrum ;

[0027] Calculate the power spectrum and convert it to dB: ;

[0028] S133, time-frequency matrix splicing: Arrange the spectrum of each frame in chronological order to form a time-frequency matrix, with the vertical axis being time and the time point calculated based on the frame shift, and the horizontal axis being frequency and the actual frequency value calculated based on the sampling rate. Furthermore, in step S200, the data of the specific time period is obtained by obtaining the data from the M1 to M2 seconds in the two-dimensional acoustic energy time-frequency image after FIR filtering, where 30≤M1≤90; 270≤M2≤360.

[0029] Furthermore, the processing and analysis of the frequency energy includes the following steps:

[0030] S210, averaging the energy values ​​of all frequency points along the matrix columns to obtain an average energy sequence with respect to the frequency;

[0031] S220, further process the average energy sequence using a sliding window with a size of M points and remove the trend, where M is any number among 7, 9, 11, 13, and 15;

[0032] S230, according to the processed frequency energy sequence, using a parameter with a set width of 10-20 as a threshold value, to find the peak value and its corresponding frequency point;

[0033] S240, for each frequency point, according to the fluctuation condition, by finding the maximum value in a certain frequency range to determine the energy fluctuation curve;

[0034] S250 , filtering, de-tilting, and DC-removing the energy fluctuation sequence to obtain a new fluctuation sequence, and then performing spectrum analysis on the sequence to obtain an energy jitter spectrum after normalization.

[0035] Furthermore, the method used for removing the trend is to remove the trend term by linear fitting.

[0036] Furthermore, when filtering the energy fluctuation sequence, a low-pass filter with an adjustable cutoff frequency is used.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] 1. The present invention has high accuracy: by comprehensively preprocessing multi-channel acoustic signals, low-frequency and high-frequency signals as well as vehicle noise and background noise are effectively suppressed, making the subsequently extracted internal wave acoustic features more accurate and reliable, and reducing the impact of noise interference on the detection results.

[0039] 2. The present invention has strong timeliness: The present invention can more efficiently extract the internal wave acoustic energy jitter characteristics from the time-frequency image. Compared with traditional methods, it improves the speed of internal wave information processing to a certain extent, helps underwater working platforms to obtain internal wave information in a timely manner, realize real-time detection and early warning, and meet the timeliness requirements in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 The acoustic time-frequency diagram with and without internal waves (the upper diagram is the case without internal waves, and the lower diagram is the case with internal waves);

[0041] Figure 2 This is the frequency response diagram of the low-frequency noise filter of the experimental measurement data;

[0042] Figure 3 These are the results before and after time-domain FIR filtering of the three channels;

[0043] Figure 4 It is the time-frequency domain analysis result of channel 1 signal after FIR processing;

[0044] Figure 5 It is the internal wave noise signal model and adaptive processing flow chart;

[0045] Figure 6 This is the test result of the noise elimination algorithm based on LMS;

[0046] Figure 7 is the time-frequency distribution image of the forward channel filtered signal (when there is no internal wave);

[0047] Figure 8 It is the analysis result of the peak and frequency points of time-frequency data;

[0048] Figure 9 This is the energy jitter spectrum analysis result obtained after normalization of channel 1 (at frequency 379);

[0049] Figure 10 This is a flow chart of the method for extracting internal wave acoustic features based on energy jitter proposed by the present invention. DETAILED DESCRIPTION

[0050] In the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," "connect," "fixed," etc. should be understood broadly. For example, they may refer to fixed connection, detachable connection, or integration; mechanical connection or electrical connection; direct connection or indirect connection through an intermediate medium; internal communication between two components or interaction between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0051] The following is a further description with reference to the accompanying drawings and specific embodiments:

[0052] 1. Proposal of acoustic energy variation characteristics in the internal wave region

[0053] 1.1 Proposal of acoustic energy variation characteristics

[0054] Internal waves are a type of nonlinear ocean wave, also known as internal solitary waves or internal solitary wave trains (hereinafter referred to as internal waves or internal solitary waves). These waves typically occur in specific areas of ocean with undulating ridges or seamounts and are excited by strong tidal currents. Once generated, they propagate forward in the open ocean as a shaped wave. These waves propagate rapidly and exhibit stable waveform characteristics in the open ocean. However, they may experience polarity reversals at shelf breaks and disperse into multiple wave trains, ultimately dissipating their energy in the shallower shelf area. Research has shown that these internal waves typically exhibit large amplitudes and turbulent currents underwater, potentially impacting the safety of both fixed and mobile underwater platforms. In addition to their large amplitudes and turbulent currents, these waves also exhibit significant surface effects. Internal ocean waves have distinct dynamic and acoustic characteristics due to their interaction with waves on the sea surface. For example, their surface flow field can cause waves to break, forming characteristics similar to those caused by wind disturbances on the sea surface. This can also change the optical and microwave scattering properties of the sea surface, and thus can be observed using optical remote sensors or synthetic aperture radars for microwave imaging. Previous research has focused on the surface or underwater ocean dynamics of internal ocean waves, while less research has been conducted on the specific acoustic characteristics of internal waves formed by the interaction between internal waves and sea surface wave fields. This present invention focuses on the analysis and extraction technology of passive acoustic characteristics of internal wave zones caused by internal wave dynamics.

[0055] 1.2 Analysis of acoustic energy change characteristics

[0056] Based on an analysis of signals obtained in the internal wave region from various hydrophone types and locations, this paper describes the "jitter" of internal wave acoustic energy around multiple frequency points in the acoustic energy time-frequency analysis image as the "energy jitter" characteristic of passive acoustics under the influence of internal waves. Signal analysis shows that different internal wave intensity, hydrophone measurement location, and channel configurations produce varying internal wave energy and jitter characteristics. Figure 1 A set of acoustic time-frequency diagrams with and without internal waves are given, where the upper sub-diagram is the acoustic time-frequency diagram without internal waves, and the lower sub-diagram is the acoustic time-frequency diagram with internal waves.

[0057] Analysis indicates that the passive acoustic energy time-frequency image of internal waves exhibits the following three main characteristics: First, the frequency locations of acoustic energy jitter are uncertain. This is related to the influence of background noise and vehicle mechanical noise in the marine environment. For example, the noise measured near the side of a ship exhibits distinct, fixed-frequency noise generated by the operation of the ship's main and auxiliary engines. However, the ship itself has little impact on the noise measured by the hydrophones carried by the autonomous underwater vehicle (AUV). However, we also noticed that the frequency locations of the jitter also vary. Second, the intensity of the energy jitter varies at different frequency locations. This characteristic in the time-frequency analysis graph is partly related to the time-frequency processing method used, such as the short-time Fourier transform or wavelet transform, which will produce significant differences. However, the fundamental reason is related to the energy of the signal itself. In fact, when converting the acquired acoustic signal into an energy time-frequency image, it can be seen that the transformed result also reflects the acoustic signal energy (such as the Fourier transform of the signal autocorrelation). This is one of the reasons why we call the passive acoustic characteristic of internal waves "energy jitter." The third manifestation is that when analyzing the time-frequency images of internal wave energy jitter over different time periods, the energy jitter changes in a consistent direction at the same moment but different frequencies. This consistency can be used to analyze the temporal variation of the jitter, which is referred to here as the energy jitter spectrum. To effectively develop a specific method for describing the characteristics of internal wave energy jitter, signal preprocessing is required, such as in the signal truncation length, filter parameters, window size, and normalization method. This preprocessing can actually highlight the characteristics of energy jitter in the time-frequency images.

[0058] 1.3 Definition of acoustic energy variation characteristics

[0059] Based on passive acoustic signals obtained in the ocean's internal wave zone, this paper proposes the concept of internal wave acoustic energy jitter characteristics, which can be used to detect the presence of internal waves. The time-varying characteristics of the jitter characteristics are referred to as "jitter spectra." Below, we design a preprocessing process for the acoustic signal, including noise cancellation and filtering, and provide specific extraction methods and techniques for obtaining this ocean internal wave acoustic energy jitter characteristic. This technical method can be used for real-time detection and early warning of the presence of such internal waves on underwater operating platforms.

[0060] like Figure 10 As shown, the present invention proposes a method for extracting internal wave acoustic features based on energy jitter, which mainly includes the following steps:

[0061] S100, preprocessing of acoustic signals in the internal wave region: performing filtering and noise cancellation processing on the multi-channel acoustic signals acquired by the hydrophone in sequence, and then obtaining a time-frequency diagram of the acoustic signals through a windowed short-time Fourier transform;

[0062] S200 , extracting internal wave acoustic energy jitter characteristics: acquiring data for a specific time period based on the acoustic signal time-frequency diagram, processing and analyzing the frequency point energy, and obtaining an energy jitter spectrum.

[0063] 2 (Step S100), Preprocessing of Internal Wave Acoustic Signals

[0064] The acoustic signals from the internal wave hydrophones described in this invention primarily consist of three main channels (more than three channels can be processed similarly): a forward measurement channel, a lateral measurement channel, and a vehicle self-noise channel (forward and lateral here refer to the front and right sides of the vehicle). Acoustic signal preprocessing involves obtaining a new forward measurement channel signal after passing it through a universal filter and two LMS adaptive filters for cancellation, thereby generating an energy time-frequency image. This process primarily involves the following three steps: First, a universal FIR bandpass filter is used to filter low- and high-frequency signals. This step requires processing signals from all channels. Second, an LMS adaptive filter is used for two cancellation filtering operations. The first filtering uses the measured vehicle self-noise channel as a reference channel (the first reference channel) for adaptive processing of the forward and lateral channel data, thereby suppressing vehicle noise through cancellation filtering. The second filtering uses the processed lateral channel signal as a reference channel to filter the forward channel again, further canceling out background noise. Third, a windowed short-time Fourier transform is used to convert the filtered signal into an acoustic signal time-frequency image.

[0065] 2.1 General FIR Bandpass Filter and Filtering Processing

[0066] Here, we choose to use a windowed FIR bandpass filter. The FIR bandpass filter designed with a Kaiser window can provide better sidelobe suppression, thus achieving the suppression of low-frequency and high-frequency signals. The Kaiser window function is shown in the following formula:

[0067] in, is the window length. It is an adjustable parameter whose shape can be controlled and can directly affect the trade-off between sidelobe attenuation and mainlobe width. It is a zero-order first-kind modified Bessel function and can reflect the smooth transition characteristics of the window function. The empirical formula is as follows The window length can be estimated using the following formula: in, is the normalized transition bandwidth (the difference between the stopband cutoff frequency and the passband cutoff frequency, divided by the sampling rate). is the stopband attenuation. This formula ensures that the transition bandwidth is minimized at a given attenuation.

[0068] The steps to design an FIR bandpass filter based on the Kaiser window include:

[0069] S111. Given filter indicators, including stopband frequency, passband frequency, stopband attenuation, and ripple parameters;

[0070] S112, calculation :according to Using the above empirical formula;

[0071] S113, confirm : Use the transition bandwidth formula to calculate the required window length;

[0072] S114, generate window function: substitute Kaiser window formula to generate time domain window;

[0073] S115. Multiply the window function by the ideal filter impulse response.

[0074] Table 1 shows the filter parameters such as stopband frequency, passband frequency, stopband attenuation and passband ripple. The frequency response of the filter is as follows: Figure 2 shown. Figure 3 The following example shows the processing results of a certain channel signal (a segment of 1 second in length is selected, and the results before and after time domain processing of three channels are shown). All measurement data are analyzed and processed using this method. From the time-frequency domain analysis results of channel 1 ( Figure 4 ), the general FIR bandpass filter can suppress low and high frequencies.

[0075] Table 1 General filter parameters

[0076] Serial number parameter Parameter value illustrate 1 Stopband frequency 0.002 Can be used to calculate transition bandwidth 2 Passband frequency 0.2 Can be used to calculate transition bandwidth 3 Stopband attenuation 60dB - 4 Passband ripple 1dB -

[0077] 2.2 Adaptive Processing of Noise Cancellation for Different Channel Signals

[0078] Considering that the lateral and forward signals received by the hydrophone on the vehicle may be affected by the noise of the vehicle itself, it is necessary to suppress the influence of the vehicle's own noise through the correlation between the signals and the use of LMS adaptive filtering. At the same time, considering the different effects of internal wave noise on the lateral and forward directions, it is assumed that the forward channel is greatly affected by the internal waves, while the lateral channel is less affected. At the same time, they are both affected by the environmental background noise. Therefore, LMS adaptive filtering can be further used to suppress the background noise to obtain enhanced internal wave influence information. Here, a time-domain internal wave signal model and an adaptive processing flow using the LMS adaptive filtering method are established, and finally the suppressed time-domain filtered signal is obtained. The LMS cancellation processing algorithm of the internal wave signal model is as follows: in , respectively represent the forward and side channels of the hydrophone on the vehicle.

[0079] The target signal output from the side channel is recorded as the reference signal, and the target signal output from the forward channel is recorded as the new signal to be processed, and the LMS cancellation processing algorithm is applied again.

[0080]

[0081] here, is the self-noise measurement value of the vehicle, is the output signal of the filter, is the forward channel signal, is the target signal after the forward channel signal is denoised and recorded. is the weight vector, is the step size parameter, To adjust the parameters. The processing flow is mainly divided into the following steps:

[0082] S121. Initialize filter parameters: Before starting adaptive filtering, initialize the filter parameters and set the initial value of the filter to 0.

[0083] S122. Read the collected multi-channel signals: including noise signals , forward and side channel signals and the signal of channel 2 .

[0084] S123 , calculating the estimated signal and the error signal by using a weighted sum of filter coefficients, and adjusting and updating the filter parameters accordingly.

[0085] S124. Repeat the above steps and continuously update the filter parameters through multiple iterations until the error of the filter output signal meets the preset requirements.

[0086] Figure 6 This is the internal wave signal model diagram. Figure 7 The following are the results of processing three channels using this method. For visual clarity, the processing steps in 2.3 are used. (a) shows the result of channel 1 offsetting channel 3; (b) shows the result of channel 1 offsetting channel 3; (c) shows the direct processing of channel 3; and (d) shows the result of channel 1 processing offsetting channel 2. Clear jitter characteristics are evident at frequencies of 1.5kHz and 3.8kHz.

[0087] 2.3 Acoustic Energy Time-Frequency Processing Based on Short-Time Fourier Transform

[0088] Time-frequency analysis is one of the important tools for processing acoustic signals. It can convert time-varying time-domain signals into two-dimensional signals of frequency and time axes, and can intuitively analyze the frequency variation characteristics of the signal at different moments in the frequency domain. The signal variation characteristics at different frequencies also reflect the change in signal energy. Here, based on the above forward channel filtering processing, a general short-time Fourier transform (STFT) processing process is further adopted to finally obtain the time-frequency distribution image of the internal wave field energy of the signal energy changing on the frequency and time axes. During the processing, the long-time non-stationary signal forward channel filtering signal is divided into short-time stationary signal segments (windowed and framed) by short-time Fourier transform, and Fourier transform is performed on each segment to obtain the joint distribution map of the signal in time and frequency - the time-frequency distribution image. The mathematical expression of short-time Fourier transform is: in is the original signal, is the window function. The processing steps are as follows:

[0089] S131, framing and windowing processing: Divide the forward channel filtered signal into pieces of length The short time frame of ; In order to reduce spectrum leakage, Hamming window processing is further adopted here; the formula is

[0090] S132, Short-time Fourier transform: Perform fast Fourier transform (FFT) on each frame of signal to obtain the complex spectrum ;

[0091] Calculate the power spectrum and convert it to dB: ;

[0092] S133, time-frequency matrix splicing: Arrange the spectrum of each frame in chronological order to form a time-frequency matrix, with the vertical axis being time and the time point calculated based on the frame shift, and the horizontal axis being frequency and the actual frequency value calculated based on the sampling rate. Figure 7The time-frequency distribution of the forward channel after FIR and LMS filtering is shown. This result shows the absence of internal waves, and no obvious signs of jitter can be seen visually.

[0093] 3 (Step S200), Extraction of Internal Wave Acoustic Energy Jitter Characteristics

[0094] Compared with the received signal before the internal wave arrives, the breaking sound formed by the interaction between the internal wave and the sea surface wave significantly increases the sound level in the area, but this increase in noise level cannot be regarded as the acoustic characteristic of the internal wave. If the time-frequency diagram of the sound energy before the internal wave arrives and when the internal wave arrives is compared, it is not difficult to find that the energy of the original frequency at multiple frequency points in the diagram has obvious left and right offsets or fluctuations, that is, the time-frequency diagram when the internal wave arrives shows the phenomenon of energy "jittering" over time near a certain frequency point. In order to extract the characteristics of the signal "jittering" phenomenon in the time-frequency diagram, the time dimension and frequency dimension phase separation processing and spectrum analysis technology are used to obtain the "jitter spectrum" diagram of the internal wave sound energy at multiple frequency points. The processing steps are as follows:

[0095] S210: Read time-frequency energy data

[0096] Based on the two-dimensional acoustic energy time-frequency image obtained after FIR filtering, the data from seconds 60 to 300 is used as the input for subsequent processing (actually a matrix). The purpose of selecting the data sequence from seconds 60 to 300 is to eliminate the convergence instability that may occur in the initial 60 seconds of the adaptive filtering sequence.

[0097] S220: Frequency energy sequence signal

[0098] The energy values ​​of all frequency points are averaged along the matrix columns to obtain the average energy sequence with respect to frequency; the energy sequence is further processed using a sliding average with a sliding window size of 11 points and the trend is removed.

[0099] S230: Calculate the peak and frequency of time-frequency data

[0100] According to the processed frequency energy sequence, the peak value and its corresponding frequency point are obtained by setting the parameter of the width as the threshold value of 10~20;

[0101] S240: Analyze the energy fluctuations of each frequency point over time

[0102] For each frequency point, according to the fluctuation situation, the energy fluctuation curve is determined by finding the maximum value in a certain frequency range.

[0103] S250: Analyze the fluctuation of sound energy at main frequency points

[0104] The energy fluctuation sequence is filtered, de-tilted and de-DC processed to obtain a new fluctuation sequence, which is then spectrally analyzed and normalized to obtain the energy jitter spectrum, that is, the acoustic energy jitter spectrum characteristics caused by internal waves.

[0105] Figure 8 It is the analysis result of the peak and frequency points of time-frequency data; Figure 9 This is the analysis result of the energy jitter spectrum after normalization. This result can be used as the main basis for determining whether internal waves exist.

[0106] In summary, this paper proposes a characteristic of acoustic energy variation caused by the interaction between the local ocean internal wave field and the sea surface wave field—the internal wave surface acoustic energy jitter characteristic—and provides techniques and methods for extracting this characteristic. This includes proposing the acoustic energy variation characteristic, i.e., the internal wave energy jitter characteristic—the spectral characteristic; implementing channel-specific filtering processing for the three-channel measurement signal; and providing methods and techniques for extracting this energy variation characteristic based on acoustic energy time-frequency images.

[0107] The present invention has high accuracy: by comprehensively preprocessing multi-channel acoustic signals, low-frequency and high-frequency signals as well as vehicle noise and background noise are effectively suppressed, making the subsequently extracted internal wave acoustic features more accurate and reliable, and reducing the impact of noise interference on the detection results.

[0108] The present invention has strong timeliness: the present invention can more efficiently extract the internal wave acoustic energy jitter characteristics from the time-frequency image. Compared with traditional methods, it improves the speed of internal wave information processing to a certain extent, helps underwater working platforms to obtain internal wave information in a timely manner, realize real-time detection and early warning, and meet the timeliness requirements in practical applications.

[0109] The above is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A method for extracting internal wave acoustic features based on energy jitter, characterized in that: The steps include: S100, preprocessing of acoustic signals in the internal wave region: performing filtering and noise cancellation processing on the multi-channel acoustic signals obtained by the hydrophone in sequence, and then obtaining a time-frequency diagram of the acoustic signals through a windowed short-time Fourier transform; the filtering processing adopts an FIR bandpass filter designed based on a Kaiser window to suppress low-frequency and high-frequency signals of the multi-channel acoustic signals; S200, extracting internal wave acoustic energy jitter characteristics: acquiring data for a specific time period based on the acoustic signal time-frequency diagram, processing and analyzing frequency point energy, and obtaining an energy jitter spectrum; The Kaiser window function is shown as follows: Where N is the window length, β is an adjustable parameter that can directly affect the trade-off between sidelobe attenuation and mainlobe width, I0(·) is the zero-order first-kind modified Bessel function, and can reflect the smooth transition characteristics of the window function. The empirical formula for parameter β is as follows The window length can be estimated using the following formula: Where Δf is the normalized transition bandwidth, A s is the stop-band attenuation; The steps to design an FIR bandpass filter based on the Kaiser window include: S111. Given filter indicators, including stopband frequency, passband frequency, stopband attenuation, and ripple parameters; S112. Calculate β: According to A s Using the above empirical formula; S113, determining N: calculating the required window length using the transition bandwidth formula; S114, generate window function: substitute Kaiser window formula to generate time domain window; S115, multiplying the window function by the ideal filter impulse response; The processing and analysis of the frequency point energy comprises the following steps: S210, averaging the energy values ​​of all frequency points along the matrix columns to obtain an average energy sequence with respect to the frequency; S220, further process the average energy sequence using a sliding window with a size of M points and remove the trend, where M is any number among 7, 9, 11, 13, and 15; S230, according to the processed frequency energy sequence, using a parameter with a set width of 10-20 as a threshold value, to find the peak value and its corresponding frequency point; S240, for each frequency point, according to the fluctuation condition, by finding the maximum value in a certain frequency range to determine the energy fluctuation curve; S250 , filtering, de-tilting, and DC-removing the energy fluctuation sequence to obtain a new fluctuation sequence, and then performing spectrum analysis on the sequence to obtain an energy jitter spectrum after normalization.

2. The method for extracting internal wave acoustic features based on energy jitter according to claim 1 is characterized in that: In step 100, the noise cancellation processing is to use an LMS adaptive filter to perform two cancellation filters. The first time, the vehicle self-noise channel is used as the reference channel, and the forward and side channel data are adaptively processed to suppress the vehicle noise. The second time, the processed side channel signal is used as the reference channel to filter the forward channel to cancel the background noise.

3. The method for extracting internal wave acoustic features based on energy jitter according to claim 2 is characterized in that: The LMS cancellation processing algorithm of the internal wave signal model is as follows: w k (n+1)=αw k (n)+μe k (n)x(n) e k (n)=d k (n)-y k (n) Where k = 1, 2, representing the forward and side channels of the hydrophone on the vehicle, respectively; x²(n)=e²(n) d(n)=e1(n) The target signal output from the side channel is recorded as the reference signal, and the target signal output from the forward channel is recorded as the new signal to be processed, and the LMS cancellation processing algorithm is applied again; w(n+1)=αw(n)+μe(n)x2(n) e(n)=d(n)-y(n) Here, y k (n) is the output signal of the filter, e k (n) is the target signal after denoising the forward channel signal; w(n) is the weight vector, μ is the step size parameter, and α is the adjustment parameter; The process of filtering by the LMS adaptive filter includes the following steps: S121, initializing filter parameters: before starting adaptive filtering, initializing the filter parameters and setting the initial value of the filter to 0; S122, reading the collected multi-channel signals: including the noise signal x(n), the forward and side channel signals d1(n), and the signal d2(n) of channel 2; S123, calculating the estimated signal and the error signal by using a weighted sum of filter coefficients, and adjusting and updating the filter parameters accordingly; S124. Repeat the above steps and continuously update the filter parameters through multiple iterations until the error of the filter output signal meets the preset requirements.

4. The method for extracting internal wave acoustic features based on energy jitter according to claim 1 is characterized in that: The mathematical expression of short-time Fourier transform is: Where x(τ) is the original signal and w(τ) is the window function; The windowed short-time Fourier transform step comprises: S131, framing and windowing processing: the forward channel filtered signal is divided into short time frames of length N, with a frame shift of H; in order to reduce spectrum leakage, Hamming window processing is further adopted here; the formula is S132, Short-time Fourier transform: Perform a fast Fourier transform (FFT) on each frame of the signal to obtain the complex spectrum X(k), calculate the power spectrum and convert it to dB: 10log 10 (|X(k)| 2 ); S133, time-frequency matrix splicing: Arrange the spectrum of each frame in chronological order to form a time-frequency matrix, with the vertical axis being time and the time point calculated based on the frame shift, and the horizontal axis being frequency and the actual frequency value calculated based on the sampling rate.

5. The method for extracting internal wave acoustic features based on energy jitter according to claim 1 is characterized in that: In the step S200 , the data of the specific time period is obtained by obtaining the data from the M1 to M2 seconds in the two-dimensional acoustic energy time-frequency image after FIR filtering, where 30≤M1≤90; 270≤M2≤360.

6. The method for extracting internal wave acoustic features based on energy jitter according to claim 1, characterized in that: The method used for removing the trend is linear fitting to remove the trend term.

7. The method for extracting internal wave acoustic features based on energy jitter according to claim 1, characterized in that: When filtering the energy fluctuation sequence, a low-pass filter with an adjustable cutoff frequency is used.

Images