Water surface vibration signal wavelet packet frequency band compression preprocessing method based on empirical mode decomposition

Abstract

The invention discloses a water surface vibration signal wavelet packet frequency band compression preprocessing method based on empirical mode decomposition, and the method comprises the steps: carrying out the wavelet packet frequency band compression preprocessing of an extracted signal, dividing the signal into different intrinsic mode functions (IMF) through adaptive noise complete set empirical mode decomposition (ICEEMDAN), and finally, the screening and reconstruction of a series of intrinsic mode functions IMFS are completed by using a correlation coefficient method, so that the separation of carrier waves and noise is realized. Wavelet packet frequency band compression mainly compresses the frequency spectrum range of a signal to be close to the frequency spectrum range of a carrier on the basis of wavelet packet transformation, and the ICEEMDAN decomposition efficiency and the mode aliasing suppression effect can be improved through the narrow frequency spectrum range. According to the method, the actual noise reduction and modal aliasing suppression effects are better than those of an ICEEMDAN method, and the extraction quality of the water surface vibration signals is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of cross-media communication, and specifically relates to a wavelet packet frequency band compression preprocessing method for water surface vibration signals based on empirical mode decomposition. Background Art

[0002] With the exploration of the ocean by people, the water-air cross-media wireless communication technology is becoming more and more important in ocean construction. Sound waves are difficult to overcome the huge impedance difference at the water-gas interface, and electromagnetic waves will be severely attenuated in water media. The failure of common transmission carriers brings challenges to water-air cross-media wireless communication.

[0003] Water-air cross-media wireless communication includes downlink communication and uplink communication. The main implementation methods of uplink communication are optical communication, acoustic electromagnetic communication and photoacoustic communication. The sound waves emitted by underwater sound sources can cause the vibration of micro-amplitude waves with the same frequency on the free water surface. By extracting the water surface vibration signals with a millimeter-wave radar, water-to-air wireless uplink communication can be realized. This method combines the advantages of acoustics and electromagnetics and overcomes the limitations of single technologies.

[0004] However, during the process of extracting water surface vibration signals with a millimeter-wave radar, the amplitude of the micro-amplitude wave vibration caused by underwater sound sources may be at the micron level or even smaller. At the same time, other interferences on the water surface will also affect the quality of the extracted signals. In addition, in order to improve the communication rate, a multi-carrier modulation method is adopted, making the frequency spectrum distribution of the extracted effective signals not concentrated, further increasing the difficulty of suppressing water surface vibration noise.

[0005] To address the above problems, a noise reduction method for water surface vibration signals based on empirical mode decomposition (EMD) is proposed. The wavelet packet frequency band compression algorithm is used to narrow the range of signal processing frequencies, and the ICEEMDAN (Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise) algorithm is used to adaptively suppress the noise of multi-carrier signals. Summary of the Invention

[0006] The purpose of the present invention is to provide a wavelet packet frequency band compression preprocessing method for water surface vibration signals based on empirical mode decomposition, which effectively improves the communication quality of signals.

[0007] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0008] A preprocessing method for wavelet packet band compression of water surface vibration signals based on empirical mode decomposition, including a power amplifier, an underwater acoustic transducer, water surface vibration, a millimeter wave radar, data preprocessing, a wavelet packet band compression algorithm, and an ICEEMDAN algorithm;

[0009] When performing transmission calculations, the information to be transmitted is first output from the host to the power amplifier, and the power amplifier drives the underwater acoustic transducer to generate modulated sound waves;

[0010] The sound waves emitted by the underwater acoustic transducer will cause transverse microwaves on the free water surface. The millimeter wave radar measures the minute vibrations on the water surface through linearly frequency-modulated electromagnetic waves. By capturing the reflected signals and mixing the reflected signals with the transmitted signals to obtain the IF signal, the water surface vibration information is extracted;

[0011] Through data preprocessing operations on the data collected by the millimeter wave radar, the vibration data of the water surface is obtained. Then, through the wavelet packet band compression algorithm, the spectral bandwidth occupied by the signal is reduced. Finally, the ICEEMDAN algorithm is used to decompose and reconstruct the signal to achieve the suppression of signal noise.

[0012] Preferably, before noise reduction, the IF signal collected by the analog-to-digital converter ADC in the system needs to be preprocessed to obtain the water surface vibration information.

[0013] Preferably, the preprocessing method for the IF signal is as follows:

[0014] First, the IF signal after ADC multiple sampling is first processed by range dimension FFT to roughly locate the water surface;

[0015] Then, the dynamic distance information between the water surface and the radar is obtained through phase extraction and phase unwrapping;

[0016] Then, according to Equation (1), it can be known that the distance R is linearly related to the initial phase of the IF signal;

[0017] Finally, the vibration frequency information of the water surface can be obtained by taking the difference of the initial phases of adjacent chirps. The vibration frequency information is shown in Equation (2):

[0018]

[0019] Preferably, the directly extracted water surface vibration signal is mixed with other echo signals and noise in the time domain. The wavelet packet band compression algorithm dynamically reduces the range of the processing content according to the frequency of the carrier signal, performs data processing, and suppresses the mode mixing of the EMD-like method;

[0020] The wavelet packet band compression algorithm performs wavelet packet decomposition on the extracted water surface vibration signal, and removes the unimportant frequency bands in the signal by setting the wavelet coefficients to zero, thereby achieving band compression.

[0021] Preferably, when the wavelet packet band compression algorithm is used for processing, the steps are as follows:

[0022] First, the wavelet packet is used to perform multi-scale decomposition on the signal S(t). The wavelet packet decomposition realizes the fine division of different frequency bands by recursively decomposing the signal.

[0023] Secondly, when performing the j-th layer of decomposition, the signal is divided into 2^j sub-bands. Each sub-band contains the information of different frequency bands of the original signal. The signal of the n-th sub-band in the j-th layer is expressed as shown in Equation (3):

[0024]

[0025] where g[x] is the low-frequency filter coefficient and h[x] is the high-frequency filter coefficient.

[0026] Then, according to the energy characteristics of each sub-band signal after wavelet packet decomposition, the sub-frequency bands of the signal are screened. The energy calculation expression is as shown in Equation (4):

[0027]

[0028] Finally, through the wavelet packet reconstruction formula, the reconstruction of the screened signal is completed to achieve the purpose of band compression, as shown in Equation (5):

[0029] W j,n (2m + k) = ∑(W j+1,2n [m]·g[k] + W j+1,2n+1 [2m - k]·h[k]) (5);

[0030] The wavelet packet basis function uses the db4 wavelet basis function, and the signal decomposition layer is 4 layers.

[0031] Preferably, during the modulated acoustic wave process, to improve the utilization rate of the frequency band and reduce the spectrum loss, the signal is modulated by using the multi-carrier method.

[0032] The ICEEMDAN algorithm adaptively decomposes the signal into a series of intrinsic mode functions (IMFs).

[0033] The ICEEMDAN algorithm is used to process the signal after wavelet packet band compression, so as to separate the carrier signal and the noise.

[0034] During the process, the ICEEMDAN algorithm suppresses mode mixing, and decomposes the carrier signal and the noise with higher stability and accuracy.

[0035] Preferably, the specific content of the ICEEMDAN algorithm is as follows:

[0036] First, the ICEEMDAN algorithm obtains the IMFs components by adding noise to the S(t) signal multiple times and performing signal decomposition;

[0037] Then, the cross-correlation operation is performed between each IMFs component and the signal S(t) to obtain the correlation coefficient;

[0038] Finally, the signal is reconstructed based on the correlation coefficient to complete noise reduction.

[0039] Preferably, the k-th IMF of signal decomposition can be expressed by the following formula (6):

[0040]

[0041] where, e j (t) is the added Gaussian white noise with zero mean and unit variance, ∈ k is the adaptive level of the noise, and M is the number of iterations;

[0042] The above r k (t) is the signal residual, which is calculated by the following formula (7):

[0043]

[0044] The cross-correlation between the signal S(t) and the intrinsic mode function (IMF) is usually used to measure the decomposition effect of the signal and the similarity of its various parts, and the correlation coefficient between each IMF component and the signal S(t) can be solved to complete the reconstruction of the signal, as shown in formula (8):

[0045]

[0046] Through the above calculation and analysis of the signal, the ICEEMDAN algorithm decomposes and reconstructs the data.

[0047] The beneficial effects of the present invention are as follows:

[0048] In the method of the present invention, the host performs noise reduction on the signal using the ICEEMDAN algorithm and the proposed method respectively, and analyzes it through the short-time Fourier transform (STFT), obtaining a more optimized noise reduction result. The noise reduction method proposed in the present invention is significantly superior to the ICEEMDAN noise reduction method. When the radar extracts the vibration signal of the water surface, the proposed noise reduction method can more effectively reduce the noise level while retaining the effective information of the signal. This method has a better noise reduction effect and can effectively improve the communication quality of the signal. Description of the Drawings

[0049] Figure 1Yes Figure 1 It is a flow chart of a signal wavelet packet band compression preprocessing method based on empirical mode decomposition.

[0050] Figure 2 It is the overall block diagram of the noise reduction method.

[0051] Figure 3 It is the signal preprocessing flow chart.

[0052] Figure 4 It is the flow chart of the wavelet packet band compression algorithm.

[0053] Figure 5 It is the flow chart of the ICEEMDAN algorithm.

[0054] Figure 6 It is a schematic diagram of the implementation platform for extracting water surface signals based on millimeter-wave radar.

[0055] Figure 7 It is a comparison chart of the effects of the proposed method and the existing method after signal noise reduction.

[0056] In the flow chart markings of the processing method, the power amplifier is 1, the underwater acoustic transducer is 2, the water surface vibration is 3, the millimeter-wave radar is 4, the data preprocessing is 5, the wavelet packet band compression algorithm is 6, and the ICEEMDAN algorithm is 7. Specific implementation mode

[0057] The present invention will be described in detail below with reference to the accompanying drawings:

[0058] It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present disclosure belongs. The following description details the embodiments of the present invention in a step-by-step manner. This description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

[0059] It should be noted that the terms used herein are only for describing specific implementation modes and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0060] Example 1

[0061] Combined with Figures 1 to 6, A wavelet packet frequency band compression preprocessing method for water surface vibration signals based on empirical mode decomposition, including a power amplifier, an underwater acoustic transducer, water surface vibration, a millimeter wave radar, data preprocessing, a wavelet packet frequency band compression algorithm, and an ICEEMDAN algorithm.

[0062] When performing transmission calculations, the information to be transmitted is first output from the host to the power amplifier, and the power amplifier drives the underwater acoustic transducer to generate modulated sound waves; the sound waves emitted by the underwater acoustic transducer will cause transverse microwaves to be generated on the free water surface. The millimeter wave radar measures the minute vibrations of the water surface through linearly frequency-modulated electromagnetic waves, captures the reflected signals, and mixes the reflected signals with the transmitted signals to obtain IF signals, thereby extracting the water surface vibration information. By performing data preprocessing operations on the data collected by the millimeter wave radar, the vibration data of the water surface is obtained, and then through the wavelet packet frequency band compression algorithm, the spectral bandwidth occupied by the signal is reduced; finally, the ICEEMDAN algorithm is used to decompose and reconstruct the signal to achieve the suppression of signal noise.

[0063] Embodiment 2

[0064] In Embodiment 1, before noise reduction, it is necessary to perform preprocessing operations on the IF signals collected by the signal digital converter ADC in the system to obtain the water surface vibration information.

[0065] The way to perform preprocessing on the IF signals is as follows:

[0066] First, the IF signals after ADC multiple sampling are first processed by distance dimension FFT to roughly locate the water surface; then, the dynamic distance information between the water surface and the radar is obtained through phase extraction and phase unwrapping; then, according to Equation (1), it can be known that the distance R has a linear relationship with the initial phase of the IF signal; finally, the vibration frequency information of the water surface can be obtained by taking the difference of the initial phases of adjacent chirps, and the vibration frequency information is shown in Equation (2):

[0067]

[0068] Embodiment 3

[0069] The sound waves emitted by the underwater acoustic transducer will cause transverse microwaves to be generated on the free water surface. For the directly extracted water surface vibration signals, other echo signals and noises will be mixed in the time domain. The wavelet packet frequency band compression algorithm dynamically reduces the range of the processing content according to the frequency of the carrier signal, performs data processing, and suppresses the mode mixing of the EMD-like method;

[0070] The wavelet packet frequency band compression algorithm decomposes the extracted water surface vibration signals by wavelet packets, and removes the unimportant frequency bands in the signals by setting the wavelet coefficients to zero, realizing frequency band compression.

[0071] Then, the wavelet packet band compression algorithm processes as follows:

[0072] First, the wavelet packet is used to perform multi-scale decomposition on the signal S(t). The wavelet packet decomposition realizes the fine division of different frequency bands by recursively decomposing the signal. Secondly, when performing j-layer decomposition, the signal is divided into 2^j sub-frequency bands, and each sub-frequency band contains the information of different frequency bands of the original signal. The nth sub-frequency band signal of the jth layer is expressed as shown in Equation (3):

[0073]

[0074] where g[x] is the low-frequency filter coefficient and h[x] is the high-frequency filter coefficient.

[0075] Then, according to the energy characteristics of each sub-frequency band signal after wavelet packet decomposition, the sub-frequency bands of the signal are screened, and the energy calculation expression is as shown in Equation (4):

[0076]

[0077] Finally, through the wavelet packet reconstruction formula, the reconstruction of the screened signal is completed to achieve the purpose of band compression, as shown in Equation (5):

[0078] W j,n (2m + k) = ∑(W j+1,2n [m]·g[k] + W j+1,2n+1 [2m - k]·h[k]) (5);

[0079] The wavelet packet basis function adopts the db4 wavelet basis function, and the signal decomposition layer number is 4 layers.

[0080] Example 4

[0081] In the process of modulating the acoustic wave in Example 1, to improve the utilization rate of the frequency band and reduce the loss of the frequency spectrum, the signal is modulated in a multi-carrier manner; the ICEEMDAN algorithm adaptively decomposes the signal into a series of intrinsic mode functions IMFs; the ICEEMDAN algorithm is used to process the signal after wavelet packet band compression, so that the carrier signal and noise are separated; during the process, the ICEEMDAN algorithm suppresses mode mixing, and the carrier signal and noise are decomposed with higher stability and accuracy.

[0082] The specific content of the ICEEMDAN algorithm is as follows:

[0083] First, the ICEEMDAN algorithm obtains the IMFs components by adding noise to the S(t) signal multiple times and signal decomposition; then, the cross-correlation operation is performed between each IMFs component and the signal S(t) to obtain the correlation coefficient; finally, the signal is reconstructed according to the correlation coefficient to complete noise reduction.

[0084] Then, the k-th IMF of signal decomposition can be expressed by the following formula (6):

[0085]

[0086] where, e j (t) is the added Gaussian white noise with zero mean and unit variance, ∈ k is the adaptive level of the noise, and M is the number of iterations;

[0087] The above r k (t) is the signal residue, which is calculated by the following formula (7):

[0088]

[0089] The cross-correlation between the signal S(t) and the Intrinsic Mode Function (IMF) is usually used to measure the decomposition effect of the signal and the similarity of its various parts. The correlation coefficient between each IMF component and the signal S(t) can be solved to complete the reconstruction of the signal, as shown in formula (8):

[0090]

[0091] Through the above calculation and analysis of the signal, the ICEEMDAN algorithm decomposes and reconstructs the data.

[0092] Example 5

[0093] For the water surface vibration signal in a certain area, during the extraction of the water surface vibration signal by the millimeter-wave radar, it is difficult to suppress the signal noise. In the related signal extraction, a water surface vibration signal noise reduction method is adopted to improve the communication quality of the signal.

[0094] During the extraction process, due to the low signal-to-noise ratio of the extracted water surface signal and the difficulty in carrier separation caused by the non-concentrated frequency distribution, a water surface vibration signal noise reduction method based on EMD is proposed. The overall process is as shown in the appendix Figure 2 as follows.

[0095] First, preprocess the data collected by the FWCM radar to obtain the vibration data of the water surface; then, reduce the spectral bandwidth occupied by the signal through the wavelet packet frequency band compression algorithm to improve the data processing effect; finally, decompose and reconstruct the signal through the ICEEMDAN algorithm to achieve the suppression of signal noise.

[0096] Before noise reduction, the IF signal collected by the ADC needs to be preprocessed to obtain the vibration information of the water surface. The processing flow is as Figure 3 shown.

[0097] The IF signal after ADC multiple sampling is first processed by range - dimension FFT to roughly locate the water surface; then the dynamic distance information between the water surface and the radar is obtained through phase extraction and phase unwrapping.

[0098] According to Equation (1), it can be known that the distance R has a linear relationship with the initial phase of the IF signal. Therefore, finally, the vibration frequency information of the water surface can be obtained by taking the difference of the initial phases of adjacent chirps, as shown in Equation (2):

[0099]

[0100] The directly extracted water - surface vibration signal will be mixed with other echo signals and noise in the time domain. Then, the wavelet packet band - compression algorithm can dynamically narrow the range of the processing content according to the frequency of the carrier signal, improve the data - processing effect, and have a certain inhibitory effect on the mode mixing of the EMD - like method. The wavelet packet band - compression algorithm decomposes the extracted water - surface vibration signal by wavelet packet and removes the unimportant frequency bands in the signal by setting the wavelet coefficients to zero, so as to achieve the effect of band compression. The algorithm implementation flowchart is as Figure 4 shown.

[0101] When the wavelet packet band - compression algorithm is processing, first use the wavelet packet to perform multi - scale decomposition on the signal S(t). The wavelet packet decomposition realizes the fine division of different frequency bands by recursively decomposing the signal. When performing j - layer decomposition, the signal is divided into 2^j sub - frequency bands, and each sub - frequency band contains the information of different frequency bands of the original signal. The signal formula of the nth sub - frequency band at the jth layer can be expressed as:

[0102]

[0103] where g[x] is the low - frequency filter coefficient and h[x] is the high - frequency filter coefficient.

[0104] According to the energy characteristics of each sub - frequency band signal after wavelet packet decomposition, the sub - frequency bands of the signal are screened. The energy calculation expression is:

[0105]

[0106] Finally, through the wavelet packet reconstruction formula, the reconstruction of the screened signal is completed to achieve the purpose of band compression, as shown in the following formula:

[0107] W j,n (2m + k)=∑(W j+1,2n [m]·g[k]+W j+1,2n+1 [2m - k]·h[k]) (5).

[0108] There are many types of wavelet packet basis functions, and the selection is crucial. The Daubechies wavelet has the characteristics of compact support and high-order vanishing moments. Therefore, in this invention, the wavelet basis function of db4 is selected, and the signal decomposition layer is 4 layers.

[0109] In order to improve the utilization rate of the frequency band and avoid wasting spectrum resources, the signal is often modulated in a multi-carrier manner. The ICEEMDAN algorithm can adaptively decompose the signal into a series of Intrinsic Mode Functions (IMFs). Using the ICEEMDAN algorithm to process the signal after wavelet packet frequency band compression can achieve a good separation effect between the carrier signal and noise. At the same time, this algorithm has better modal aliasing suppression ability, higher decomposition stability and accuracy than the EMD and CEEMDAN algorithms.

[0110] The ICEEMDAN algorithm obtains the IMF components by adding noise to the S(t) signal multiple times and decomposing it; then, the cross-correlation operation is performed between each IMF component and the signal S(t) to obtain the correlation coefficient; finally, the signal is reconstructed according to the correlation coefficient to complete noise reduction. The algorithm flow block diagram is as Figure 5 shown.

[0111] The k-th IMF of signal decomposition can be expressed by the following formula:

[0112]

[0113] where, e j (t) is the added Gaussian white noise with zero mean and unit variance, ∈ k is the adaptive level of the noise, M is the number of iterations, r k (t) is the signal residual, which can be calculated by formula (7).

[0114]

[0115] The cross-correlation between the signal S(t) and the Intrinsic Mode Function (IMF) is usually used to measure the decomposition effect of the signal and the similarity of its various parts. The correlation coefficient between each IMF component and the signal S(t) can be solved to complete the reconstruction of the signal, as shown in formula (8):

[0116]

[0117] Example 6

[0118] In order to verify the noise reduction effect of the proposed method, in this example, a platform for extracting the water surface vibration signal based on a millimeter-wave radar was built, as Figure 6As shown, it is tested in the pool. During the whole experimental process, the host outputs 2FSK multi-carrier modulation information, which is amplified by a power amplifier to drive the transducer to generate modulated sound waves, thus causing slight vibrations on the water surface. Finally, the millimeter-wave radar extracts the water surface vibration information and transmits the IF signal to the host.

[0119] The host performs noise reduction on the signal using the ICEEMDAN algorithm and the proposed method respectively, and analyzes it through the short-time Fourier transform (STFT) to obtain the noise reduction results as Figure 7 shown. By comparing Figure 7 the noise reduction results of the two signals in Figure 7 the right), it can be seen that the proposed noise reduction method ( Figure 7 right) is significantly better than the ICEEMDAN noise reduction method ( Figure 7 left). When the radar extracts the vibration signal of the water surface, the proposed noise reduction method can more effectively reduce the noise level while retaining the effective information of the signal. The final results show that the method proposed in the present invention has a better noise reduction effect and can effectively improve the communication quality of the signal.

Claims

1. A wavelet packet band compression preprocessing method for water surface vibration signals based on empirical mode decomposition, characterized in that: Including power amplifier, underwater acoustic transducer, water surface vibration, millimeter wave radar, data pre-processing, wavelet packet band compression algorithm, ICEEMDAN algorithm; When processing and calculating, the information to be transmitted is first output to the power amplifier through the host, and the power amplifier drives the underwater acoustic transducer to generate modulated sound waves; The sound waves emitted by the hydroacoustic transducer will cause the free water surface to generate lateral microwaves. The millimeter wave radar measures the tiny vibrations of the water surface through linear frequency-sweeping electromagnetic waves, captures the reflected signal, and mixes the reflected signal with the transmitted signal to obtain the IF signal, thereby extracting the water surface vibration information; By performing data preprocessing operations on the data collected by the millimeter-wave radar, the vibration data of the water surface is obtained, and then the spectrum bandwidth occupied by the signal is reduced through the wavelet packet band compression algorithm. Finally, the signal is decomposed and reconstructed through the ICEEMDAN algorithm to suppress signal noise.

2. The method for preprocessing the wavelet packet frequency band compression of water surface vibration signals based on empirical mode decomposition according to claim 1 is characterized in that: Before noise reduction, the IF signal collected by the signal digital converter ADC in the system needs to be pre-processed to obtain the vibration information of the water surface.

3. The method for preprocessing the wavelet packet frequency band compression of water surface vibration signals based on empirical mode decomposition according to claim 2 is characterized in that: The IF signal is pre-processed as follows: First, the IF signal after ADC multi-sampling is first processed by range-dimensional FFT to roughly locate the water surface; Then, the dynamic distance information between the water surface and the radar is obtained through phase extraction and phase unwrapping; Then, according to formula (1), it can be seen that the distance R is linearly related to the initial phase of the IF signal; Finally, the vibration frequency information of the water surface can be obtained by subtracting the initial phases of adjacent chirps. The vibration frequency information is shown in formula (2):

4. The method for preprocessing the wavelet packet frequency band compression of water surface vibration signals based on empirical mode decomposition according to claim 1 is characterized in that: The directly extracted water surface vibration signal will be mixed with other echo signals and noise in the time domain. The wavelet packet band compression algorithm dynamically reduces the scope of processing content according to the frequency of the carrier signal, performs data processing, and suppresses the modal aliasing of the EMD-like method. The wavelet packet frequency band compression algorithm decomposes the extracted water surface vibration signal into wavelet packets and removes unimportant frequency bands in the signal by setting the wavelet coefficients to zero, thereby achieving frequency band compression.

5. The method for preprocessing the wavelet packet frequency band compression of water surface vibration signals based on empirical mode decomposition according to claim 4 is characterized in that: The wavelet packet band compression algorithm is processed as follows: Firstly, the signal S(t) is decomposed into multiple scales using wavelet packets. Wavelet packet decomposition recursively decomposes the signal to achieve fine division of different frequency bands. Secondly, when performing j-layer decomposition, the signal is divided into 2j sub-bands, each of which contains information of different frequency bands of the original signal. The n-th sub-band signal of the j-th layer is expressed as shown in formula (3): Where g[x] is the low-frequency filter coefficient and g[x] is the high-frequency filter coefficient; Then, the sub-bands of the signal are screened according to the energy characteristics of each sub-band signal after wavelet packet decomposition. The energy calculation expression is as follows: Finally, the signal after screening is reconstructed through the wavelet packet reconstruction formula to achieve the purpose of frequency band compression, as shown in formula (5): W j,n (2m+k)=∑(W j+1,2n [m]·g[k]+W j+1,2n+1 [2m-k]·h[k]) (5); The wavelet packet basis function adopts the wavelet basis function of db4, and the number of signal decomposition layers is 4.

6. The method for preprocessing the wavelet packet frequency band compression of water surface vibration signals based on empirical mode decomposition according to claim 1 is characterized in that: In the process of modulating sound waves, the utilization rate of frequency band is improved, the loss of spectrum is reduced, and the signal is modulated by multi-carrier method; The ICEEMDAN algorithm adaptively decomposes the signal into a series of intrinsic mode functions (IMFs); The ICEEMDAN algorithm is used to process the signal after wavelet packet band compression, so that the carrier signal and noise can be separated. During the process, the ICEEMDAN algorithm performs modal aliasing suppression and decomposes the carrier signal and noise with greater stability and accuracy.

7. The method for preprocessing the wavelet packet frequency band compression of water surface vibration signals based on empirical mode decomposition according to claim 6 is characterized in that: The specific content of the ICEEMDAN algorithm is: First, the ICEEMDAN algorithm obtains the IMFs components by performing multiple noise addition and signal decomposition on the S(t) signal; Then, the correlation coefficient is obtained by performing cross-correlation operation between each IMFs component and the signal S(t); Finally, the signal is reconstructed according to the correlation coefficient to complete the noise reduction.

8. The method for preprocessing the wavelet packet frequency band compression of water surface vibration signals based on empirical mode decomposition according to claim 7, characterized in that: The kth IMF of the signal decomposition can be expressed by the following formula (6): Among them, e j (t) is the added Gaussian white noise with zero mean and unit variance, ∈ k is the adaptive level of noise, M is the number of iterations; The above k (t) is the signal residual, which is calculated by the following formula (7): The cross-correlation between the signal S(t) and the intrinsic mode function (IMF) is usually used to measure the decomposition effect of the signal and the similarity of its parts. The correlation coefficient between each IMF component and the signal S(t) can be solved to complete the reconstruction of the signal, as shown in formula (8): The ICEEMDAN algorithm calculates and analyzes the signal as mentioned above, and the data is decomposed and reconstructed.

Images