Deep sea underwater sound data compression method and device based on FPGA (Field Programmable Gate Array)

Abstract

The invention relates to the technical field of sound wave data processing, in particular to a deep sea underwater sound data compression method and device based on an FPGA, and the method comprises the steps: obtaining the energy distribution of IMF waves according to the mode decomposition of deep sea underwater sound data, and carrying out the segmentation processing, and obtaining each segment of sound data; according to the amplitude values of any two sections of sound data and the change similarity condition, obtaining sound similarity measurement; obtaining information concentration according to the sound data of the same time period on all different IMF waves and the sound similarity measurement of different sound data on the same IMF wave; and selecting a wavelet basis function group in a wavelet transformation process according to the information concentration, quantifying and encoding wavelet coefficients after all IMF wave wavelet transformation, and obtaining a compression result of the deep sea underwater sound data. The compression quality of the deep sea underwater sound data can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of acoustic wave data processing, and in particular to a method and device for compressing deep sea water acoustic wave data based on FPGA. Background Art

[0002] Deep-sea acoustic data refers to the sound wave signal data collected by sonar devices and other equipment in the deep-sea environment. Due to the huge amount of data generated by deep-sea exploration and the limited communication bandwidth in the deep-sea environment, the storage and transmission costs of data are high, which requires the deep-sea acoustic data obtained by detection to be compressed so that it can be transmitted to the corresponding ship or platform.

[0003] At present, the commonly used method for compressing deep sea acoustics is the lossless compression method based on dictionary coding. However, due to the presence of various deep sea noises distributed at different frequencies in deep sea acoustics, when the method of directly denoising and then directly compressing is used, the complex noise distribution will cover up some data features, which will cause the loss of these feature data during the denoising process, affecting the compression quality of deep sea acoustic data. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of this application is to provide a deep sea water acoustic data compression method and device based on FPGA. The technical solutions adopted are as follows: In a first aspect, an embodiment of the present application provides a deep sea water acoustic data compression method based on FPGA, the method comprising the following steps: S1, obtain deep sea acoustic data; S2, decompose the deep sea water acoustic data modally to obtain IMF waves, perform segment processing according to the energy distribution of each IMF wave, and obtain each segment of sound data; for any two segments of sound data, obtain the sound similarity measurement of the two segments of sound data according to the similarity of the amplitude values ​​and changes of the two segments of sound data; S3, performing threshold analysis on the sound similarity measures between the sound data of the same time period on all different IMF waves to obtain the sound residual similarity of each segment of sound data; combining the sound similarity measures and sound residual similarities of each segment of sound data with all other sound data on the same IMF wave to obtain the noise characteristic value of each segment of sound data; S4, according to the proportion of all sound data on each IMF wave and the noise characteristic value, the information concentration of each IMF wave is obtained; according to the information concentration of each IMF wave, the wavelet basis function group of each IMF wave in the wavelet transform process is selected; according to the wavelet basis function group of each IMF wave, each IMF wave is subjected to wavelet transformation to obtain the wavelet coefficient, the wavelet coefficients of all IMF waves are quantized and encoded, and the compression result of deep sea water acoustic data is obtained.

[0005] Furthermore, the modal decomposition of deep sea water acoustic data to obtain IMF waves is performed, and segment processing is performed according to the energy distribution of each IMF wave to obtain sound data of each segment, including: The deep sea water acoustic data is modally decomposed to obtain a preset number of IMF waves; each segment of the IMF wave is framed to obtain each frame signal; each frame signal is judged in turn using the energy threshold method to filter out each segment of sound data.

[0006] Furthermore, the energy threshold method is used to determine each frame signal in turn, and each segment of sound data is screened, including: Calculate the energy of each frame signal; obtain the mode of the energy of all frame signals, and use one quarter of the mode as the energy threshold; record each frame signal with energy higher than the energy threshold as each segment of sound data.

[0007] Furthermore, for any two segments of sound data, the sound similarity measurement of the two segments of sound data is obtained according to the similarity of the amplitude values ​​and changes of the two segments of sound data, including: , In the formula, Indicates sound data With sound data Sound similarity measure of; Indicates sound data Middle Amplitude value at the moment; Indicates sound data Middle Amplitude value at the moment; Indicates sound data With sound data DTW distance; is an exponential function with a natural constant as base; is the number of moments collected in each segment of sound data.

[0008] Furthermore, the method for obtaining the sound residual similarity of each segment of sound data includes: For any IMF wave, any segment of sound data on the IMF wave is recorded as the target data, and the sound data on other IMF waves with the same time period as the target data is recorded as the co-location data of the target data; the number of sound similarity measurements between the target data and all its co-location data that is greater than or equal to a preset similarity threshold is obtained and recorded as the similarity number; the ratio of the similarity number to the number of all IMF waves is taken as the residual sound similarity of the target data.

[0009] Furthermore, the method for obtaining the noise characteristic value of each segment of sound data includes: The average value of the sound similarity measure between any one segment of sound data and all other sound data on the same IMF wave is obtained; and the product of the average value and the sound residual similarity of the any one segment of sound data is used as the noise characteristic value of the any one segment of sound data.

[0010] Furthermore, the method for obtaining the information density of each IMF wave comprises: The information density of an IMF wave is recorded as , ; In the formula, Indicates the length of each segment of sound data; Indicates sound data The noise characteristic value of Indicates The total amount of sound data contained in an IMF wave; Indicates The total length of an IMF wave.

[0011] Further, the wavelet basis function group of each IMF wave in the wavelet transform process is selected according to the information density of each IMF wave, including: Obtaining a normalized value of the information concentration of each IMF wave; rounding up a preset multiple of the normalized value to obtain an information level of each IMF wave; Randomly select a number of wavelet basis function groups; the number of wavelet basis function groups is consistent with the number of information levels, and each wavelet basis function group has two orthogonal wavelet bases; All wavelet basis function groups are sorted in order from small to large according to the scale of the wavelet basis, and the serial number of each wavelet basis function group is recorded as the scale level of each wavelet basis function group; Among all the scale levels of the wavelet basis function groups, the wavelet basis function groups with the same information level values ​​as the IMF waves are used as the wavelet basis function groups of the IMF waves in the wavelet transform process.

[0012] Further, the wavelet transformation of each IMF wave according to the wavelet basis function group of each IMF wave is performed to obtain the wavelet coefficient, and the wavelet coefficients of all IMF waves are quantized and then encoded to obtain the compression result of the deep sea water acoustic data, which specifically includes: Each IMF wave is decomposed and transformed using its wavelet basis function group in the wavelet transform process to obtain wavelet coefficients; the wavelet coefficients are uniformly quantized; the uniformly quantized wavelet coefficients are encoded using entropy coding to obtain encoded data of each IMF wave; the encoded data of all IMF waves are combined into a coding table for deep sea water acoustic data.

[0013] In a second aspect, an embodiment of the present application further provides an FPGA-based deep sea water acoustic data compression device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements any one of the steps of the above-mentioned FPGA-based deep sea water acoustic data compression method when executing the computer program.

[0014] This application has at least the following beneficial effects: This application analyzes the problem that some non-noise key information is lost when the existing technology is used to compress deep-sea acoustic data after denoising, and realizes efficient compression of deep-sea acoustic signals by utilizing the high-speed parallel processing capability and flexible programmable characteristics; first, the original acoustic data is decomposed to obtain the intrinsic mode functions of different frequencies, which is conducive to more efficient and accurate search for noise data that may exist in the acoustic data; further analyze the relationship between the similarity of different sound data segments in each IMF wave and the noise content, obtain the noise characteristic value, and more accurately evaluate the analysis of how much noise each IMF wave may contain; further combine the sound data proportion and noise characteristic value on each IMF wave to obtain the information concentration, evaluate the information content of each IMF wave, so as to select different wavelet bases for each IMF wave to perform wavelet transform on the IMF wave, which can compress the noise data more accurately without deleting the noise data, and effectively retain other key information and data features. Finally, the quantization technology is combined to quantize and encode many wavelet coefficients and IMF wave coefficients to complete the compression of deep-sea acoustic data, which can effectively improve the compression quality of deep-sea acoustic data in deep-sea operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 A flowchart of the steps of a deep sea water acoustic data compression method based on FPGA provided in one embodiment of the present application; Figure 2 A block diagram of a wavelet basis function group provided for one embodiment of the present application. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0018] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0019] The specific scheme of the deep sea water acoustic data compression method and device based on FPGA provided by the present application is described in detail below with reference to the accompanying drawings.

[0020] See also Figure 1 , which shows a flowchart of a deep sea water acoustic data compression method based on FPGA provided by an embodiment of the present application, the method comprising the following steps: S1, obtain deep sea acoustic data.

[0021] The implementation scenario of this application is to compress the deep-sea water acoustic data collected and fed back when positioning or detecting deep-sea terrain, organisms and other objects in the deep sea by means of sound waves and sonar devices. A deep-sea distributed acoustic receiving device is set on deep-sea submersibles including unmanned underwater submersibles and other equipment, and then the submersible is lowered to the deep-sea target position to collect deep-sea water acoustic data. Then, the deep-sea water acoustic data is input into the field programmable gate array, and the water acoustic signal is pre-processed through its built-in digital signal processing module, and the data received by multiple water acoustic collection devices are integrated and divided to obtain several segments of integrated deep-sea water acoustic data. The collection time of each segment of deep-sea water acoustic data is 1 hour, and the deep-sea water acoustic data is normalized.

[0022] S2, modally decompose the deep sea water acoustic data to obtain IMF waves, and perform segmented processing according to the energy distribution of each IMF wave to obtain each segment of sound data; for any two segments of sound data, the sound similarity measurement of the two segments of sound data is obtained according to the similarity of the amplitude values ​​and changes of the two segments of sound data.

[0023] In the deep sea, when collecting deep sea water acoustics, we often encounter various noises caused by the movement of seawater, the effect of wind and atmosphere on the sea surface, the movement or melting of water layers, changes in seabed geological structures, the sounds made by marine organisms, and man-made sources. These noises are either low-frequency or high-frequency, widely distributed, and appear at irregular times. Usually, component signals at different frequencies are obtained through EMD decomposition; then the amount of noise on each component is analyzed separately. Among them, the EMD empirical mode decomposition algorithm is a well-known technology, and the specific implementation process will not be repeated.

[0024] Considering that the noise is more on a certain component signal, or less on a certain component, when performing wavelet transform on these components, wavelet bases of different sizes can be selected; that is, for components with more noise, a larger wavelet can be selected to improve the compression rate and reduce the proportion of noise in the final transmitted information. Conversely, selecting a small wavelet base can retain more detailed information in the underwater sound. Finally, the coefficients in the EMD decomposition and wavelet transform are encoded through quantization technology to complete the compression of deep sea water acoustic data.

[0025] Specifically, this embodiment decomposes deep sea water acoustic data by the EMD empirical mode decomposition algorithm to obtain a preset number of intrinsic modal components, i.e., IMF waves. In this embodiment, 10 IMF waves are decomposed. In order to further analyze the IMF waves obtained by EMD decomposition, a suitable wavelet basis is selected; first, the possibility of each small segment of sound data belonging to noise is analyzed. Each segment of IMF wave is framed, and the preset frame size is 20 milliseconds, and each frame overlaps 50%. Frame processing can help extract local features of the signal, and the signal characteristics in each frame can be considered to be relatively stable. After that, each frame signal is judged in turn by the energy threshold method, and frames less than the energy threshold are eliminated to filter out each segment of sound data. In this embodiment, the energy of each frame signal is first calculated, and then the mode of the energy of all frame signals is selected, and one quarter of it is taken as the energy threshold. The calculation of the energy of the signal is a well-known technology and will not be repeated. Each frame signal with energy higher than the energy threshold is recorded as each segment of sound data.

[0026] In the deep sea environment, there are many noises, which are widely distributed in many frequencies, and the amount of noise at different frequencies is also different; the noise sources are usually fixed, so these noises have certain similarities, which are reflected in volume and tone, and the relevant features of each noise also have certain repetition in the time domain. Based on this, this embodiment analyzes the possibility that each small segment of sound data belongs to noise.

[0027] First, for any two sound data, the smaller the difference in volume between the two sound data at many moments and the higher the similarity in the changes in the pitch of the two sound data, the more similar the two sound data are; considering that this process only considers the similarity at a certain moment, the similarity of the two sounds is obtained by adding up the similarities at many moments. Based on this, the similarity between the two sound data is calculated to evaluate the similarity between the two sounds, and the calculation formula is: , In the formula, Indicates sound data With sound data Sound similarity measure of; Indicates sound data Middle Amplitude value at the moment; Indicates sound data Middle Amplitude value at the moment; Indicates sound data With sound data DTW distance; is an exponential function with a natural constant as base; is the number of moments collected in each segment of sound data.

[0028] in, It is used to describe the similarity between two sound data in terms of time domain and pitch. The smaller the DTW distance, the higher the similarity between the two. It is used for inverse normalization to describe the similarity between two sounds in terms of quantity and pitch. The closer the amplitudes of the two sounds at each moment are, the higher the similarity between the two sounds, and the greater the sound similarity measure obtained.

[0029] S3, performing threshold analysis on the sound similarity measure between the sound data of the same time period on all different IMF waves to obtain the sound residual similarity of each segment of sound data; combining the sound similarity measure and sound residual similarity of each segment of sound data with all other sound data on the same IMF wave to obtain the noise characteristic value of each segment of sound data.

[0030] According to the characteristics of deep-sea sound, each small sound segment is analyzed to determine whether it belongs to deep-sea noise. When a submersible is sailing in the deep sea, the sound data collected is relatively complex. However, after EMD decomposition of various sounds belonging to noise, a part of the sound segment will still remain in each component, which leads to the continuous repetition of this noise in other components. Based on this, the residual similarity of the sound of the target segment is constructed.

[0031] First, the sound data segments corresponding to each sound data segment on other IMF waves are obtained, and the corresponding time points of the two sound data segments on different IMFs are the same.

[0032] For any IMF wave, any segment of sound data on the IMF wave is recorded as the target data, and the sound data on other IMF waves with the same time period as the target data is recorded as the co-location data of the target data; the number of sound similarity measurements between the target data and all its co-location data that is greater than or equal to a preset similarity threshold is obtained and recorded as the similarity number; the ratio of the similarity number to the number of all IMF waves is taken as the residual sound similarity of the target data.

[0033] Among them, by presetting the similarity threshold After screening, the sound residual similarity obtained describes the sound fluctuation similarity between the target data and the data on other IMF waves in the same time period.

[0034] As deep-sea noise, its repeatability is higher than other sounds. The similarity between sound data on the same IMF wave is further combined to judge the possibility that it belongs to noise. Specifically, the average value of the sound similarity measurement between any segment of sound data and other segments of sound data on the same IMF wave is obtained; the product of the average value and the sound residual similarity of the any segment of sound data is used as the noise characteristic value of the any segment of sound data.

[0035] The noise characteristic value is used to describe the possibility that the data is deep-sea noise. The noise characteristic value is obtained by combining the sound similarity measurement of the sound data with other sound data on the same IMF wave. Specifically, the average value of the sound similarity measurement of any sound data segment and all other sound data on the same IMF wave is obtained; the product of the average value and the sound residual similarity of the any sound data segment is used as the noise characteristic value of the any sound data segment.

[0036] S4, according to the proportion of all sound data on each IMF wave and the noise characteristic value, the information concentration of each IMF wave is obtained; according to the information concentration of each IMF wave, the wavelet basis function group of each IMF wave in the wavelet transform process is selected; according to the wavelet basis function group of each IMF wave, each IMF wave is subjected to wavelet transformation to obtain the wavelet coefficient, the wavelet coefficients of all IMF waves are quantized and encoded, and the compression result of deep sea water acoustic data is obtained.

[0037] Since the bandwidth for transmitting information is limited in the deep-sea working environment, and in order to retain more deep-sea acoustic data, it is necessary to distinguish the ratio of valid data to noise data in each IMF wave. In deep-sea acoustic data, noise of each frequency exists. For each IMF wave of the corresponding frequency, the less noise it contains, the more valid acoustic data the IMF wave contains, that is, the more valid acoustic data the corresponding component contains, the more important the IMF wave is. Based on this, this embodiment evaluates the importance of each IMF wave: , In the formula, Indicates The information density of an IMF wave; Indicates the length of each segment of sound data; Indicates sound data The noise characteristic value of Indicates The total amount of sound data contained in an IMF wave; Indicates The total length of an IMF wave.

[0038] Information density is used to evaluate the importance of valid hydroacoustic data contained in the IMF wave; the noise eigenvalue indicates the possibility that the sound data contains valid data. The less noise, the more information it contains; the greater the proportion of sound data in the IMF wave, the smaller the degree of noise interference; by taking the inverse normalized value of the noise eigenvalue as the weight and performing weighted averaging on the duration of the sound data, the greater the information density obtained, the more valid information is contained in the IMF wave.

[0039] Since there are many kinds of noise in the deep sea, there are different degrees of noise on each IMF wave after decomposition. For IMF waves with more noise, they contain relatively less effective information, so their compression degree can be increased to give up transmission bandwidth and storage space for those with more effective information. Therefore, in order to select a suitable wavelet basis for each IMF wave during wavelet transformation, this embodiment first constructs the wavelet basis level to be selected according to the information concentration of each IMF wave: , In the formula, Indicates The level of information content of each IMF wave; Indicates Information density of an IMF wave Use the maximum and minimum normalization to get the normalized result; Represents the ceiling function.

[0040] Among them, the information level is used to describe the level of the wavelet basis that needs to be selected; The information density of the first IMF wave is projected onto all the information densities to observe their relative importance. Then, the information density is amplified tenfold and rounded up to obtain the The information level of an IMF wave ranges from 1 to 10, that is, 10 levels. The closer to 10, the more effective information the IMF wave contains.

[0041] In order to adaptively adjust the compression ratio of wavelet transform, each IMF wave is transformed by biorthogonal wavelet. Therefore, multiple different wavelet basis function groups are selected in advance, each of which has two orthogonal wavelet bases, and the number of wavelet basis function groups is consistent with the number of information levels, that is, 10 levels.

[0042] In order to select a larger wavelet basis for the IMF wave with higher information content when decomposing the IMF wave, so that the noise data can be compressed to a greater extent, and vice versa, a smaller wavelet basis is selected to retain more detailed features in the deep sea acoustic data, which can effectively remove the noise. Therefore, the scale level corresponding to the wavelet basis function with a smaller scale is higher. In this embodiment, all wavelet basis function groups are encoded from 1 to 10 in order from small to large according to the scale of the wavelet basis, and encoded as the scale level of the wavelet basis function group. Among the scale levels of all wavelet basis function groups, the wavelet basis function group with the same information level value of each IMF wave is used as the wavelet basis function group of each IMF wave in the wavelet transform process; that is, the wavelet basis function group of the same level can be selected for each IMF wave for transformation.

[0043] The selection box diagram of the wavelet basis function group is as follows: Figure 2 shown.

[0044] The wavelet basis functions required for each IMF wave are obtained above, and each IMF wave is decomposed by these wavelet basis functions to obtain the corresponding wavelet coefficients. The information level of each IMF wave is unique, and the wavelet basis level selected when performing wavelet transform is also uniquely determined. In addition, since each IMF wave is decomposed and split from deep sea water acoustic data. Therefore, the deep sea water acoustic data after noise reduction is obtained by transforming and superimposing the IMF waves. Therefore, in this embodiment, the wavelet coefficients of each IMF wave are quantized and encoded to obtain the encoded data of each IMF wave; the encoded data of all IMF waves are composed of the encoding table of deep sea water acoustic data to complete the compression of deep sea water acoustic data. Among them, the method for quantizing the wavelet coefficients is uniform quantization, and the method for compressing the quantized wavelet coefficients is entropy coding. The specific implementation process is a well-known technology, and this application will not repeat it.

[0045] Based on the same inventive concept as the above method, an embodiment of the present application further provides an FPGA-based deep sea water acoustic data compression device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, the steps of any one of the above-mentioned FPGA-based deep sea water acoustic data compression methods are implemented.

[0046] Through the above description of the implementation method in combination with the accompanying drawings, technical personnel in the relevant field can understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0047] The above contents are only specific implementation methods of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be covered by the protection scope of the present application.

Claims

1. A deep sea water acoustic data compression method based on FPGA, characterized in that: The method comprises the following steps: S1, obtain deep sea acoustic data; S2, decompose the deep sea water acoustic data modally to obtain IMF waves, perform segment processing according to the energy distribution of each IMF wave, and obtain each segment of sound data; for any two segments of sound data, obtain the sound similarity measurement of the two segments of sound data according to the similarity of the amplitude values ​​and changes of the two segments of sound data; S3, performing threshold analysis on the sound similarity measures between the sound data of the same time period on all different IMF waves to obtain the sound residual similarity of each segment of sound data; combining the sound similarity measures and sound residual similarities of each segment of sound data with all other sound data on the same IMF wave to obtain the noise characteristic value of each segment of sound data; S4, according to the proportion of all sound data on each IMF wave and the noise characteristic value, the information concentration of each IMF wave is obtained; according to the information concentration of each IMF wave, the wavelet basis function group of each IMF wave in the wavelet transform process is selected; according to the wavelet basis function group of each IMF wave, each IMF wave is subjected to wavelet transformation to obtain the wavelet coefficient, the wavelet coefficients of all IMF waves are quantized and encoded, and the compression result of deep sea water acoustic data is obtained.

2. The deep sea water acoustic data compression method based on FPGA as claimed in claim 1, characterized in that: The modal decomposition of deep sea water acoustic data to obtain IMF waves is performed, and segment processing is performed according to the energy distribution of each IMF wave to obtain each segment of sound data, including: The deep sea water acoustic data is modally decomposed to obtain a preset number of IMF waves; each segment of the IMF wave is framed to obtain each frame signal; each frame signal is judged in turn using the energy threshold method to filter out each segment of sound data.

3. The deep sea water acoustic data compression method based on FPGA as claimed in claim 2, characterized in that: The energy threshold method is used to determine each frame signal in turn, and each segment of sound data is obtained by screening, including: Calculate the energy of each frame signal; obtain the mode of the energy of all frame signals, and use one quarter of the mode as the energy threshold; record each frame signal with energy higher than the energy threshold as each segment of sound data.

4. The deep sea water acoustic data compression method based on FPGA as claimed in claim 1, characterized in that: For any two segments of sound data, the sound similarity measurement of the two segments of sound data is obtained according to the similarity of the amplitude values ​​and changes of the two segments of sound data, including: ， In the formula, Indicates sound data With sound data Sound similarity measure of; Indicates sound data Middle Amplitude value at the moment; Indicates sound data Middle Amplitude value at the moment; Indicates sound data With sound data DTW distance; is an exponential function with a natural constant as base; is the number of moments collected in each segment of sound data.

5. The deep sea water acoustic data compression method based on FPGA as claimed in claim 1, characterized in that: The method for obtaining the sound residual similarity of each segment of sound data comprises: For any IMF wave, any segment of sound data on the IMF wave is recorded as the target data, and the sound data on other IMF waves with the same time period as the target data is recorded as the co-location data of the target data; the number of sound similarity measurements between the target data and all its co-location data that is greater than or equal to a preset similarity threshold is obtained and recorded as the similarity number; the ratio of the similarity number to the number of all IMF waves is taken as the residual sound similarity of the target data.

6. The deep sea water acoustic data compression method based on FPGA as claimed in claim 1, characterized in that: The method for obtaining the noise characteristic value of each segment of sound data comprises: The average value of the sound similarity measure between any one segment of sound data and all other sound data on the same IMF wave is obtained; and the product of the average value and the sound residual similarity of the any one segment of sound data is used as the noise characteristic value of the any one segment of sound data.

7. The deep sea water acoustic data compression method based on FPGA as claimed in claim 1, characterized in that: The method for obtaining the information concentration of each IMF wave comprises: The information density of an IMF wave is recorded as , ; In the formula, Indicates the length of each segment of sound data; Indicates sound data The noise characteristic value of Indicates The total amount of sound data contained in an IMF wave; Indicates The total length of an IMF wave.

8. The deep sea water acoustic data compression method based on FPGA as claimed in claim 1, characterized in that: The wavelet basis function group of each IMF wave in the wavelet transform process is selected according to the information density of each IMF wave, including: Obtaining a normalized value of the information concentration of each IMF wave; rounding up a preset multiple of the normalized value to obtain an information level of each IMF wave; Randomly select a number of wavelet basis function groups; the number of wavelet basis function groups is consistent with the number of information levels, and each wavelet basis function group has two orthogonal wavelet bases; All wavelet basis function groups are sorted in order from small to large according to the scale of the wavelet basis, and the serial number of each wavelet basis function group is recorded as the scale level of each wavelet basis function group; Among all the scale levels of the wavelet basis function groups, the wavelet basis function groups with the same information level values ​​as the IMF waves are used as the wavelet basis function groups of the IMF waves in the wavelet transform process.

9. The deep sea water acoustic data compression method based on FPGA as claimed in claim 1, characterized in that: The method of performing wavelet transformation on each IMF wave according to the wavelet basis function group of each IMF wave to obtain wavelet coefficients, and quantizing and encoding the wavelet coefficients of all IMF waves to obtain the compression result of deep sea water acoustic data specifically includes: Each IMF wave is decomposed and transformed using its wavelet basis function group in the wavelet transform process to obtain wavelet coefficients; the wavelet coefficients are uniformly quantized; the uniformly quantized wavelet coefficients are encoded using entropy coding to obtain encoded data of each IMF wave; the encoded data of all IMF waves are combined into a coding table for deep sea water acoustic data.

10. A deep sea water acoustic data compression device based on FPGA, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the deep sea water acoustic data compression method based on FPGA as claimed in any one of claims 1 to 9 are implemented.

Images