Data transmission method based on compressed sensing

Abstract

The invention discloses a data transmission method based on compressed sensing, and the method comprises the steps: enabling an acquisition module to transmit seismic data to a single-chip microcomputer for caching through acquisition interruption, carrying out the discrete wavelet transformation through the acquisition interruption after the caching data is fully written, reading a chaotic matrix parameter from an eMMC chip, observing a wavelet transformation coefficient, storing an observation value into a MiniSEED data package, and transmitting the observation value to a storage module; in a main program, a file system is called to transmit a compressed data packet to an upper computer through 4G, after the upper computer obtains a compressed data file, a wavelet transform coefficient is analyzed by using a compressed sensing reconstruction algorithm, and then original seismic data is recovered through wavelet inverse transformation, and the method has the advantages of being high in compression speed, low in algorithm complexity, good in signal fidelity and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of high-precision real-time data transmission, and in particular to a data transmission method based on compressed sensing. Background Art

[0002] In earthquake-prone areas, digital seismographs are widely used to monitor the dynamic changes of roads when earthquakes occur and to take safety precautions in advance. Therefore, real-time data transmission is particularly important. The analog signals collected by the seismograph are converted into one-dimensional digital signals through AD. During the data collection process, the data needs to be compressed and uploaded, and the compression algorithm needs to have a high speed. Therefore, there is an urgent need for technical solutions that can improve the performance and speed of the compression algorithm. Summary of the invention

[0003] The purpose of the present invention is to solve the above-mentioned deficiencies in the prior art and to provide a data transmission method based on compressed sensing with fast compression speed, low algorithm complexity and good signal fidelity.

[0004] The technical solution adopted by the present invention to solve its technical problem is:

[0005] A data transmission method based on compressed sensing, characterized by adopting the following method:

[0006] Step S1: The acquisition module sends the seismic data to the internal cache of the single-chip microcomputer through the acquisition interrupt;

[0007] Step S2: When the cache data is full, discrete wavelet transform is performed through acquisition interruption;

[0008] Step S3: Read out the chaotic matrix parameters from the eMMC chip to observe the wavelet transform coefficients;

[0009] Step S4: storing the observed values ​​into the MiniSEED data package;

[0010] Step S5: In the main program, the file system is called to transmit the compressed data packet to the host computer via 4G;

[0011] Step S6: After the host computer obtains the compressed data file, it uses the compressed sensing reconstruction algorithm to parse out the wavelet transform coefficients, and then restores the original seismic data through the inverse wavelet transform;

[0012] Data compression and reconstruction based on compressed sensing can achieve high-precision real-time data transmission and reduce the time and bandwidth required for data transmission.

[0013] The specific method of discrete wavelet transform in step S2 of the present invention is:

[0014] Step S2.1: Input Lv, X S[N], Hi[WL], Lo[WL], where X S [N] is a single-precision floating-point seismic data sequence of length N, Hi[WL] is a wavelet basis high-frequency coefficient of length WL, Lo[WL] is a wavelet basis low-frequency coefficient of length WL, Lv is the number of wavelet decomposition layers, and continue to execute step S2.2;

[0015] Step S2.2: Define i=1, cA0=X S , lA0=N, where i is the i-th wavelet decomposition layer, i∈{1,2,...,Lv}, execute step S2.3;

[0016] Step S2.3: Input sequence cA i-1 Extend the edge to get where cA i-1 is the input sequence for the i-th layer wavelet decomposition, and executes step S2.4;

[0017]

[0018] Step S2.4: Decompose the i-th layer of wavelet;

[0019]

[0020] for n i =1,3,5,……,lA i ;

[0021]

[0022]

[0023] pass Discrete convolution with Hi[WL] and Lo[WL] to obtain the high-frequency coefficient cD of the i-th layer decomposition i [n i ] and the low frequency coefficient cA i [n i ], execute step S2.5;

[0024] Step S2.5: count up, i=i+1, execute step S2.6;

[0025] Step S2.6: Determine whether i is greater than Lv, if yes, jump to step S2.7, if no, jump to step S3;

[0026] Step S2.7: Output the length set of the wavelet transform coefficient sequence Lw = [lD1, lD2, ..., lDL v ,lAL v ], the length of the i-th (i∈{1,2,...,Lv}) layer decomposition is:

[0027]

[0028] Where lA0=N,lD i =lA i , i∈{1,2,...,Lv}, the total length of wavelet transform coefficients Nl is the sum of the elements of the length set Lw, and the calculation formula is:

[0029]

[0030] Output wavelet transform coefficient sequence Cw = [cD1, cD2, ..., cD Lv , cA Lv ] T , wavelet decomposition is completed.

[0031] The method of chaotic matrix linear observation in step S3 of the present invention is:

[0032] Construct N l ×M-dimensional Logistic chaos matrix Φ={ε1,ε2,...,ε Μ} T , and stored in the Logistic.txt file of the eMMC, where ε i , i∈{1,2,...,M} is the row vector of Φ. After the wavelet transform process is completed, the row vectors of the Logistic chaotic matrix are taken to the operation buffer row by row, and the inner product operation is performed with the wavelet transform coefficient c respectively to obtain the chaotic matrix linear observation value y.

[0033] The wavelet of the present invention adopts the db3 wavelet of the Daubechies series.

[0034] In step S6 of the present invention, the specific method of using the compressed sensing reconstruction algorithm to parse the wavelet transform coefficients and then recovering the original seismic data through the wavelet inverse transform is:

[0035] Step S6.1: Input MiniSEED data packets D1, D2, D3..., D i , execute step S6.2;

[0036] Step S6.2: Determine whether the encoding format is X, if yes, execute step S6.3, if no, execute step S6.11;

[0037] Step S6.3: define k=1, and execute step S6.4;

[0038] Step S6.4: i<{k, k+1, k+2}%3, execute step S6.5;

[0039] Step S6.5: Determine whether the file has been read completely, if so, execute step S6.12, if not, execute step S6.6;

[0040] Step S6.6: Parse D i Get four consecutive observations y i1 ,y i2 ,y i3 ,y i4 , i∈{1,2,3}, execute step S6.7;

[0041] Step S6.7: gOMP algorithm reconstructs the approximation c of the wavelet transform coefficients i1 ,c i2 ,c i3 ,c i4 , i∈{1,2,3}, execute step S6.8;

[0042] Step S6.8: Inverse wavelet transform to obtain x i1 ,x i2 ,x i3 ,x i4 , execute step S6.9;

[0043] Step S6.9: Based on the inverse wavelet transform, x i1 ,x i2 ,x i3 ,x i4 , the original time series Xi={x i1 ,x i2 ,x i3 ,x i4}, i∈{1,2,3}, execute step S6.10;

[0044] Step S6.10: define k=k+3, and jump to step S6.4;

[0045] Step S6.11: if it is determined to be uncompressed format data, execute step S6.12;

[0046] Step S6.12: Complete data decompression and end.

[0047] The beneficial effects of the present invention are: data compression and reconstruction based on compressed sensing can achieve high-precision real-time data transmission and reduce the time and bandwidth required for data transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is the overall compression and reconstruction process based on compressed sensing of the present invention.

[0049] Figure 2 It is the wavelet transform input parameter configuration table of the present invention.

[0050] Figure 3 It is a principle flow chart of multi-layer discrete wavelet transform of the present invention.

[0051] Figure 4 It is a flow chart of the data decompression host computer program of the present invention.

[0052] Figure 5 It is the waveform diagram of the experimental seismic data of the present invention.

[0053] Figure 6 It is the simulation performance index of the present invention. Figure 7 This is a comparison chart of the compressed sensing signal reconstruction effect of the present invention. DETAILED DESCRIPTION

[0054] The present invention is described below in conjunction with the accompanying drawings and embodiments.

[0055] As attached Figure 1-4 As shown, a data transmission method based on compressed sensing adopts the following method:

[0056] Step S1: The acquisition module sends the seismic data to the internal cache of the single-chip microcomputer through the acquisition interrupt;

[0057] Step S2: When the cache data is full, discrete wavelet transform is performed through acquisition interruption;

[0058] Step S2.1: Input Lv, X S [N], Hi[WL], Lo[WL], where X S [N] is a single-precision floating-point seismic data sequence of length N, Hi[WL] is a wavelet basis high-frequency coefficient of length WL, Lo[WL] is a wavelet basis low-frequency coefficient of length WL, Lv is the number of wavelet decomposition layers, and continue to execute step S2.2;

[0059] Step S2.2: Define i=1, cA0=X S , lA0=N, where i is the i-th wavelet decomposition layer, i∈{1,2,...,Lv}, execute step S2.3;

[0060] Step S2.3: To ensure that the input sequence cA is i-1 The discrete convolution operation with the wavelet basis coefficients will not cause memory leaks. For p = 2ni-k+1, the input sequence cA i-1 Extend the edge to get Execute step S2.4;

[0061]

[0062] Step S2.4: Decompose the i-th layer of wavelet;

[0063]

[0064] for n i =1,3,5,……,lA i ;

[0065]

[0066]

[0067] pass Discrete convolution with Hi[WL] and Lo[WL] to obtain the high-frequency coefficient cD of the i-th layer decomposition i [n i ] and the low frequency coefficient cA i [n i ], execute step S2.5;

[0068] Step S2.5: count up, i=i+1, execute step S2.6;

[0069] Step S2.6: Determine whether i is greater than Lv, if yes, jump to step S2.7, if no, jump to step S3;

[0070] Step S2.7: Output the length set of the wavelet transform coefficient sequence Lw = [lD1, lD2, ..., lDL v ,lAL v ], the length of the i-th (i∈{1,2,...,Lv}) layer decomposition is:

[0071]

[0072] Where lA0=N,lD i =lA i , i∈{1,2,...,Lv}, the total length of wavelet transform coefficients Nl is the sum of the elements of the length set Lw, and the calculation formula is:

[0073]

[0074] Output wavelet transform coefficient sequence Cw = [cD1, cD2, ..., cD L , cA Lv ] T Wavelet decomposition is completed;

[0075] Step S3: Read out the chaotic matrix parameters from the eMMC chip to observe the wavelet transform coefficients;

[0076] Construct N l ×M-dimensional Logistic chaos matrix Φ={ε1,ε2,...,εΜ} T , and stored in the Logistic.txt file of the eMMC, where ε i , i∈{1,2,...,M} is the row vector of Φ. After the wavelet transform process is completed, the row vectors of the Logistic chaotic matrix are taken row by row to the operation buffer area, and the inner product operation is performed with the wavelet transform coefficient c respectively to obtain the chaotic matrix linear observation value y;

[0077] In this embodiment, Nl=1030, compression ratio CR=N / M=4, observation length M=252 is obtained, and an Nl×M dimensional Logistic chaotic matrix is ​​constructed;

[0078] Step S4: storing the observed values ​​into the MiniSEED data package;

[0079] Step S5: In the main program, the file system is called to transmit the compressed data packet to the host computer via 4G;

[0080] Step S6: After the host computer obtains the compressed data file, it uses the compressed sensing reconstruction algorithm to parse out the wavelet transform coefficients, and then restores the original seismic data through the inverse wavelet transform;

[0081] Step S6.1: Input MiniSEED data packets D1, D2, D3..., D i , execute step S6.2;

[0082] Step S6.2: Determine whether the encoding format is X, if yes, execute step S6.3, if no, execute step S6.11;

[0083] Step S6.3: define k=1, and execute step S6.4;

[0084] Step S6.4: i<{k, k+1, k+2}%3, execute step S6.5;

[0085] Step S6.5: Determine whether the file has been read completely, if so, execute step S6.12, if not, execute step S6.6;

[0086] Step S6.6: Parse D i Get four consecutive observations y i1 ,y i2 ,y i3 ,y i4 , i∈{1,2,3}, execute step S6.7;

[0087] Step S6.7: gOMP algorithm reconstructs the approximation c of the wavelet transform coefficients i1 ,c i2 ,c i3 ,ci4 , i∈{1,2,3}, execute step S6.8;

[0088] Step S6.8: Inverse wavelet transform to obtain x i1 ,x i2 ,x i3 ,x i4 , execute step S6.9;

[0089] Step S6.9: Based on the inverse wavelet transform, x i1 ,x i2 ,x i3 ,x i4 , the original time series Xi={x i1 ,x i2 ,x i3 ,x i4}, i∈{1,2,3}, execute step S6.10;

[0090] Step S6.10: define k=k+3, and jump to step S6.4;

[0091] Step S6.11: if it is determined to be uncompressed format data, execute step S6.12;

[0092] Step S6.12: Complete data decompression and end;

[0093] Data compression and reconstruction based on compressed sensing can achieve high-precision real-time data transmission and reduce the time and bandwidth required for data transmission.

[0094] In this embodiment, a segment of seismic data in the form of 32-bit AD values ​​collected by the seismometer at a sampling rate of 4000SPS is converted into a voltage amplitude (full scale ±2.5V), and 16128 sampling points are intercepted as the experimental data set DATA. Figure 5 This is the waveform diagram drawn by the data DATA-Z of the Z channel of the data segment. It is divided into 16 segments for subsequent experiments. The sequence length L = 1008, the data type is float32, and the number of iterations of the greedy algorithm K = M / 4. Through experimental analysis, the wavelet uses the db3 wavelet of the Daubechies series to perform 5-layer wavelet decomposition, and the wavelet transform coefficients are observed with the Logistic chaotic matrix with a compression ratio of CR = 4. When decompressing, the gOMP algorithm is used to reconstruct the observed values. The compression effect of the scheme is verified by simulation software. Under the conditions of the above parameters, the segmented experimental data set DATA-Z is compressed and decompressed respectively. By calculating its PSNR and RMSE of the 32-bit AD value signal, the time complexity of the algorithm is obtained using simulation software. The performance indicators are as follows: Figure 6As shown (the wavelet transform coefficients are obtained by transforming the 32-bit binary AD value, and the original sequence is converted into an amplitude sequence, unit: V), Figure 6 and Figure 7 It can be seen that when the compressed sensing algorithm is used to compress 32-bit seismic data, the PSNR can reach 81.786 dB and the RMSE is 3.4963×10 5 , which is equivalent to the noise level at μV level, so the signal fidelity is good, the algorithm complexity is low, and the compression rate is sufficient for real-time compression of seismic data.

Claims

1. A data transmission method based on compressed sensing, characterized in that: Use the following method: Step S1: The acquisition module sends the seismic data to the internal cache of the single-chip microcomputer through the acquisition interrupt; Step S2: When the cache data is full, discrete wavelet transform is performed through acquisition interruption; Step S3: Read out the chaotic matrix parameters from the eMMC chip to observe the wavelet transform coefficients; Step S4: storing the observed values ​​into the MiniSEED data package; Step S5: In the main program, the file system is called to transmit the compressed data packet to the host computer via 4G; Step S6: After the host computer obtains the compressed data file, it uses the compressed sensing reconstruction algorithm to parse out the wavelet transform coefficients, and then restores the original seismic data through the inverse wavelet transform.

2. A data transmission method based on compressed sensing according to claim 1, characterized in that: The specific method of discrete wavelet transform in step S2 is: Step S2.1: Input Lv, X S [N], Hi[WL], Lo[WL], where X S [N] is a single-precision floating-point seismic data sequence of length N, Hi[WL] is a wavelet basis high-frequency coefficient of length WL, Lo[WL] is a wavelet basis low-frequency coefficient of length WL, Lv is the number of wavelet decomposition layers, and continue to execute step S2.2; Step S2.2: Define i=1, cA0=X S , lA0=N, where i is the i-th wavelet decomposition layer, i∈{1,2,...,Lv}, execute step S2.3; Step S2.3: Input sequence cA i-1 Extend the edge to get where cA i-1 is the input sequence for the i-th layer wavelet decomposition, and executes step S2.4; Step S2.4: Decompose the i-th layer of wavelet; for n i =1,3,5,……,lA i ; pass Discrete convolution with Hi[WL] and Lo[WL] to obtain the high-frequency coefficient cD of the i-th layer decomposition i [n i ] and the low frequency coefficient cA i [n i ], execute step S2.5; Step S2.5: count up, i=i+1, execute step S2.6; Step S2.6: Determine whether i is greater than Lv, if yes, jump to step S2.7, if no, jump to step S3; Step S2.7: Output the length set of the wavelet transform coefficient sequence Lw = [lD1, lD2, ..., lD Lv ,lA Lv ], the length of the i-th (i∈{1,2,...,Lv}) layer decomposition is: Where lA0=N,lD i =lA i , i∈{1,2,...,Lv}, the total length of wavelet transform coefficients N l is the sum of the elements of the length set Lw, and the calculation formula is: Output wavelet transform coefficient sequence Cw = [cD1, cD2, ..., cD Lv ,cA Lv ] T Wavelet decomposition is complete.

3. A data transmission method based on compressed sensing according to claim 1 or 2, characterized in that: The method of chaotic matrix linear observation in step S3 is: Construct N l ×M-dimensional Logistic chaos matrix Φ={ε1,ε2,...,ε Μ } T , and stored in the Logistic.txt file of the eMMC, where ε i , i∈{1,2,...,M} is the row vector of Φ. After the wavelet transform process is completed, the row vectors of the Logistic chaotic matrix are taken to the operation buffer row by row, and the inner product operation is performed with the wavelet transform coefficient c respectively to obtain the chaotic matrix linear observation value y.

4. A data transmission method based on compressed sensing according to claim 1 or 2, characterized in that: The wavelet adopts the db3 wavelet of the Daubechies series.

5. A data transmission method based on compressed sensing according to claim 1 or 2, characterized in that: In step S6, the specific method of using the compressed sensing reconstruction algorithm to parse the wavelet transform coefficients and then restore the original seismic data through the wavelet inverse transform is: Step S6.1: Input MiniSEED data packets D1, D2, D3..., D i , execute step S6.2; Step S6.2: Determine whether the encoding format is X, if yes, execute step S6.3, if no, execute step S6.11; Step S6.3: define k=1, and execute step S6.4; Step S6.4: i<{k, k+1, k+2}%3, execute step S6.5; Step S6.5: Determine whether the file has been read completely, if so, execute step S6.12, if not, execute step S6.6; Step S6.6: Parse D i Get four consecutive observations y i1 ,y i2 ,y i3 ,y i4 , i∈{1,2,3}, execute step S6.7; Step S6.7: gOMP algorithm reconstructs the approximation c of the wavelet transform coefficients i1 ,c i2 ,c i3 ,c i4 , i∈{1,2,3}, execute step S6.8; Step S6.8: Inverse wavelet transform to obtain x i1 ,x i2 ,x i3 ,x i4 , execute step S6.9; Step S6.9: According to the inverse wavelet transform, x i1 ,x i2 ,x i3 ,x i4 , the original time series Xi={x i1 ,x i2 ,x i3 ,x i4 }, i∈{1,2,3}, execute step S6.10; Step S6.10: define k=k+3, and jump to step S6.4; Step S6.11: if it is determined to be uncompressed format data, execute step S6.12; Step S6.12: Complete data decompression and end.

Images