A digital channelized sensing method based on MWC

Abstract

The application relates to a kind of digital channeling sensing methods based on MWC, belonging to the field of digital signal processing. It includes: integer decimation operation is carried out to signal, frequency mixing, then through corresponding delay polyphase filter bank, frequency mixing;After that, it is divided and summed according to interval, and DFT is carried out after summation;Finally, it is sent into CTF module to obtain joint support set;According to joint support set, the original signal is restored. By introducing MWC structure, the digital channeling structure is compressed, the subsequent processing rate is greatly reduced, the whole system only needs to save part of narrowband data to restore the original wideband sparse data, and detection of signal is not needed, only the signal needs to satisfy sparse priori.

Description

Technical Field

[0001] This invention belongs to the field of digital signal processing and relates to a digital channelization sensing system based on the modulated wideband converter (MWC) concept, which optimizes the digital channelization receiving scheme by utilizing the modulated wideband converter structure. Background Technology

[0002] Broadband digital channelized receivers typically receive non-cooperative radar pulse signals, whose frequencies change rapidly over short periods. This necessitates real-time analysis of sampled data to achieve spectrum sensing capabilities in highly dynamic environments. However, existing broadband digital sensing technologies suffer from a problem with the digital channelization model: channel partitioning doesn't truly reduce data throughput; the overall transmission and processing rate remains unchanged, only allowing for distribution to multiple computing units. Therefore, to further reduce data processing load, a more reasonable digital sensing receiver scheme is needed to ultimately achieve rapid detection of non-cooperative signals.

[0003] In real-world applications, signals typically exhibit sparse distribution characteristics across different transform domains. Compressed sensing theory proposes that if the sampling structure or content of a signal satisfies sparsity constraints, a CS sampling system can reconstruct the original signal with high probability. In 2006, the U.S. Defense Advanced Research Projects Agency (DARPA) proposed the concept of an analog-to-digital converter (AIC) to promote the practical application of CS sampling theory. Typical AIC structures include random demodulators (RDs) and modulation broadband converters.

[0004] The MWC structure overcomes the high input analog bandwidth requirement of the ADC in the MCS and RD structures by utilizing the characteristics of channelization and the concept of blind signal recovery. By introducing a periodic random modulation signal, MWC modulates the original broadband spectrum to a specific baseband position, and then filters out the baseband signal through a low-pass filter. By utilizing the characteristics of different signals within the broadband spectrum at the baseband, the original signal is recovered, enabling low-rate acquisition and processing of broadband signals. Summary of the Invention

[0005] The technical problem to be solved by this invention is:

[0006] When processing digital signals acquired by broadband high-speed ADCs, traditional digital channelization processing schemes divide the high-speed data stream into multiple narrowband data streams through multi-phase decimation and filtering to facilitate real-time processing by subsequent computing units. However, this does not substantially reduce the data throughput required for subsequent processing; it remains equal to the original data throughput. To further reduce the data processing load, this invention provides a digital channelization sensing method based on MWC (Modulation-Driven Conversion). The MWC structure can modulate and compress broadband signals, requiring only a small amount of narrowband data streams at the backend to complete the sensing and reconstruction of the entire broadband spectrum. By improving the digital channelization architecture through the MWC structure concept, the data rate of subsequent data transmission and processing is reduced, achieving the reconstruction of broadband information from narrowband processing.

[0007] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0008] A digital channelization sensing method based on MWC, characterized in that it includes:

[0009] The band-limited signal x(t) is sampled at the Nyquist sampling rate to obtain the digital signal x(n);

[0010] Performing integer decimation on a digital signal x(n) yields a new data sequence x. l (m), for the data sequence x l (m) is mixed to obtain data x l (k);

[0011] Data x l (k) Obtain data x through a multiphase filter bank with corresponding delay. l (k), for data x l (k) performs frequency mixing to obtain the signal. The polyphase filter bank is the l-th polyphase component of the prototype filter h0(n), and the normalized cutoff frequency of the prototype filter is 1 / L.

[0012] Signal Divide the data into intervals D and sum the results. Then perform an M-point DFT to obtain the signal y. m (k); where D and M satisfy L = DM;

[0013] Signal y m (k) Input the joint support set into the CTF module According to the joint support set Restore the original signal.

[0014] A further technical solution of the present invention: the integer decimation operation on the digital signal x(n) yields a new data sequence x. l (m), specifically

[0015] Take the l-th value from every L consecutive data points to obtain new data:

[0016] x l (m)=x(n)| n＝mL+l

[0017] In the formula, L is the decimation coefficient, which is a positive integer, and l is the decimation phase.

[0018] A further technical solution of the present invention: the data sequence x l (m) is mixed to obtain data x l (k), specifically

[0019] x l (k)=x l (k)((-1) (L-1)k p l (k))

[0020] Where, p l (k) represents the l-th polyphase component sequence of the random modulation sequence P(n). The length of the random modulation sequence is N, and the value is randomly taken as +1 or -1. The equivalent symbol sequence periodic frequency is f. p =f nyq / N,f nyq This is the Nyquist sampling rate.

[0021] A further technical solution of the present invention: the data x l (k) performs frequency mixing to obtain the signal. Specifically:

[0022]

[0023] A further technical solution of the present invention: the signal Divide the data into intervals D and sum the results. Then perform an M-point DFT to obtain the signal y. m (k), specifically:

[0024]

[0025] A further technical solution of the present invention: the signal y m (k) Input the joint support set into the CTF module Specifically:

[0026] Using the CTF module, the problem can be transformed into an MMV problem, thus enabling the estimation of the support set.

[0027]

[0028] Where, vector y[n] = [y1[n],…,ym [n] T The i-th element in nT represents the value in nT. s The sampled value of the i-th channel of the system at time t; any one of them satisfies Q = VV H The matrix V can be called the frame of y(f), i.e., V = E Q D Q D Q It is composed of the eigenvalues ​​of matrix Q The diagonal matrix formed, E Q It is composed of eigenvectors Given an m×m dimensional matrix composed of permutations; if the equation y(f)=Az(f) has a unique joint sparse solution z(f) and matrix A satisfies the RIP condition, then V=AU also has a unique joint sparse solution. support set Support set of z(f) Maintain consistency; utilize the CS reconstruction algorithm for the joint sparse matrix By making an estimate, the joint support set of z(f) can be obtained.

[0029] A further technical solution of the present invention: using the orthogonal matching pursuit algorithm to solve the V=AU problem decomposed by the CTF module, specifically:

[0030] Let the residual R0 = V, the sparse vector U = 0, and the support set The number of iterations l = 1;

[0031] Finding the best-matching atom: where R l-1 Let a represent the residual after the last iteration. k Let <,> represent the k-th column of matrix A, and <,> represent the inner product.

[0032]

[0033] Update support set

[0034]

[0035] Update residuals

[0036]

[0037]

[0038] Until the iteration stops.

[0039] A further technical solution of the present invention: the method based on the joint support set To restore the original signal, specifically:

[0040]

[0041] A further technical solution of the present invention includes:

[0042] For each sequence z l Performing L-fold interpolation on [n], 1≤l≤L, increases the sequence rate to the sampling rate of the signal x(t), yielding...

[0043]

[0044] The interpolated sequence Through the ideal interpolation filter h l After [n], time-domain modulation and summation are performed to reconstruct the sampling sequence of the original signal.

[0045]

[0046] A computer system is characterized by comprising: one or more processors, and a computer-readable storage medium for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method described above.

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

[0048] This invention provides a digital channelization sensing method based on MWC (Multi-Wideband Convergence). By introducing the MWC structure to compress the digital channelization structure, the subsequent processing rate is greatly reduced. The entire system only needs to save a portion of narrowband data to recover the original wideband sparse data, and it does not require signal detection, only that the signal satisfies the sparse prior. It has the following three advantages:

[0049] 1. By using a digital channelization structure, a single-channel MWC structure can be equivalent to a multi-channel structure through frequency shifting. This greatly simplifies the implementation of the original multi-channel MWC structure. Furthermore, through polyphase decomposition, the high-speed data stream is split into multiple low-speed polyphase data streams, making it more suitable for hardware implementation.

[0050] 2. By introducing the wideband modulation concept from the MWC structure, digital signals can be compressed using random modulation sequences. This eliminates the need for prior signal detection, allowing for direct data compression and significantly reducing the amount of data requiring subsequent processing, thus facilitating data transmission.

[0051] 3. Compared to the original digital channelization structure, this structure requires less computational resources by merging and simplifying the original structure's channels. This is mainly reflected in the fact that only D adders are added, while the number of DFT operation points is reduced by a factor of D. Attached Figure Description

[0052] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0053] Figure 1 This is the digital channelized sensing structure of the method of the present invention;

[0054] Figure 2 This is a flowchart of the process of the method of the present invention;

[0055] Figure 3 (a) is the frequency domain pattern of the original signal in the image domain in an embodiment of the present invention; (b) is the frequency domain pattern of the original signal in the frequency hopping interval from the 475th to the 495th time in the image domain in an embodiment of the present invention.

[0056] Figure 4 (a) is the frequency domain pattern of the image domain recovered signal in an embodiment of the present invention; (b) is the frequency domain pattern of the image domain recovered signal in the frequency hopping interval from the 475th to the 495th time in an embodiment of the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0058] This invention provides a digital channelization sensing method based on MWC, such as... Figure 2 As shown, it includes the following steps:

[0059] Step 1: Assume x(t) is a space L f The Fourier transform of a continuous signal with finite energy is:

[0060]

[0061] If signal x(t) is a band-limited signal, its spectral range is F = [-1 / 2T]. nyq 1 / 2T nyq The corresponding Nyquist sampling rate is f. nyq =1 / T nyq X(f) has only K disjoint subbands, and the bandwidth of each subband does not exceed B.

[0062] At Nyquist sampling rate f nyq When x(t) is sampled, the digital signal received by the system is x(n).

[0063] Step 2: Perform integer extraction on the original sampling sequence, which involves taking the l-th value from every L consecutive data points to obtain new data:

[0064] x l (m)=x(n)| n＝mL+l

[0065] In the formula, L is the decimation coefficient, which is a positive integer, and l is the decimation phase.

[0066] Step 3: Mix the new data sequence to obtain x. l (k).

[0067] x l (k)=x l (k)((-1) (L-1)k p l (k))

[0068] p l (k) represents the l-th polyphase component sequence of the random modulation sequence P(n). The random modulation sequence has a length of N, and the values ​​are randomly selected (+1 or -1). The equivalent symbol sequence periodic frequency is f. p =f nyq / N. This process can be represented as...

[0069] P(n)=α n ∈{+1,-1}

[0070] For a random modulation sequence P(n), the number of symbols P in each period must satisfy the following condition:

[0071]

[0072] Step 4: Data x from Step 3 r (k) Through a polyphase filter bank with corresponding delays. Define H l (m)=h0(n)|n=mL+l is the l-th polyphase component of the prototype filter h0(n), and the normalized cutoff frequency of the prototype filter is 1 / L.

[0073]

[0074] Step 5: Convert the data x obtained in Step 4 l (k) performs frequency mixing to obtain the signal.

[0075]

[0076] Step 6: Transfer the data obtained in Step 5 Divide the data into intervals D and sum the results, then perform an M-point DFT. D and M satisfy L = DM.

[0077]

[0078] Step 7: After the above steps, the received signal is now equivalent to:

[0079]

[0080]

[0081] Its frequency domain expression is:

[0082]

[0083] At this time, the perception matrix A m×m The remaining rows are transformed into cyclic shifts of c1(l). Calculate the discrete Fourier series of the modulation sequence and construct the sensing matrix A based on this series.

[0084]

[0085]

[0086] Step 8: Take the sampling sequence y[n] = {y1[n],...,y1[n]} obtained in Step 6 and convert it into a single sample. m [n]} is sent to the CTF module.

[0087] The processing steps for the CTF module are as follows:

[0088] Using the CTF module, the problem can be transformed into an MMV problem, thus enabling the estimation of the support set.

[0089]

[0090] Where, vector y[n] = [y1[n],…,y m [n] T The i-th element in nT represents the value in nT. s The sampled value of the i-th channel of the system at time t. Any sampled value satisfying Q = VV H The matrix V can be called the frame of y(f), such as V = E Q D Q D Q It is composed of the eigenvalues ​​of matrix Q The diagonal matrix formed, E Q It is composed of eigenvectors Given an m×m dimensional matrix composed of permutations. If the equation y(f)=Az(f) has a unique joint sparse solution z(f), and matrix A satisfies the RIP condition, then V=AU also has a unique joint sparse solution. support set Support set of z(f) To maintain consistency, it is only necessary to use the CS reconstruction algorithm on the joint sparse matrix. By making an estimate, the joint support set of z(f) can be obtained.

[0091] Step 9: Solve the V=AU problem decomposed by the CTF module in Step 8 using the Orthogonal Matching Pursuit algorithm to obtain the joint support set.

[0092] The specific process of the orthogonal matching pursuit algorithm is as follows:

[0093] Step 10: Let the residual R0 = V, the sparse vector U = 0, and the support set... The number of iterations l = 1;

[0094] Step 11: Find the best matching atom: where R l-1 Let a represent the residual after the last iteration. k Let <,> represent the k-th column of matrix A, and <,> represent the inner product.

[0095]

[0096] Step 12: Update the support set

[0097]

[0098] Step 13: Update residuals

[0099]

[0100]

[0101] Step 14: Verify whether the iteration stopping condition is met: If the condition is met, i.e., l≥K or ||R l If ||2≤ε, then proceed to step 15; otherwise, let l=l+1 and return to step 11.

[0102] Step 15: Based on the joint support set obtained in Step 13 Restore the original signal.

[0103]

[0104] Step 16: Since the rate of z[n] is the same as the rate of y[n], it is necessary to process each sequence z... l Performing L-fold interpolation on [n], 1≤l≤L, increases the sequence rate to the sampling rate of the signal x(t), yielding...

[0105]

[0106] The interpolated sequence Through the ideal interpolation filter h l After [n], time-domain modulation and summation are performed to reconstruct the sampling sequence of the original signal.

[0107]

[0108] To enable those skilled in the art to better understand the present invention, the present invention will be described in detail below with reference to specific embodiments.

[0109] Example:

[0110] Step 1: Assume the signal is a wideband signal with a signal-to-noise ratio of 10dB and a bandwidth of 0.01GHz, which hops from 0.1GHz to 4.9GHz 1000 times in steps of 0.048GHz, with each hop lasting for (1024*256 / 5e) seconds. 9 The original signal frequency hopping pattern is as follows: Figure 3 As shown in (a), assume that the number of snapshots per sampling is 1024. Using the Nyquist sampling rate f... nyq =10e 9 Hz samples x(t), and the digital signal received by the system at this time is x(n).

[0111] Step 2: Perform integer extraction on the original sampling sequence, which involves taking the l-th value in the range [0, 255] from every consecutive L = 256 data points to obtain new data:

[0112] x l (m)=x(n)| n＝mL+l

[0113] Where L is the decimation coefficient, which is a positive integer, and l is the decimation phase.

[0114] Step 3: Mix the new data sequence to obtain x. l (k).

[0115] x l (k)=x l (k)((-1) (L-1)k p l (k))

[0116] Where, p l (k) represents the l-th polyphase component sequence ∈ [0, 255] of the random modulation sequence P(n). The random modulation sequence P(n) comes from a sequence of length N = 256 sequences generated by MATLAB with seed rng(100) and following a uniform distribution α. ik For a sequence ∈ {+1,-1}, the equivalent symbol sequence has a periodic frequency of f. p =f nyq / N. This process can be represented as...

[0117] P(n)=α n ∈{+1,-1}

[0118] Step 4: Data x from Step 3 r (k) Through a polyphase filter bank with corresponding delays. Define H l (m)=h0(n)|n=mL+l is the l-th polyphase component of the prototype filter h0(n), and the normalized cutoff frequency of the prototype filter is 1 / L.

[0119]

[0120] Step 5: Convert the data x obtained in Step 4 l (k) performs frequency mixing to obtain the signal.

[0121]

[0122] Step 6: Transfer the data obtained in Step 5 Take D = 64 points at intervals M = 4, sum them, and then perform an M-point DFT. D and M satisfy L = DM.

[0123]

[0124] Step 7: After the above steps, the received signal is now equivalent to:

[0125]

[0126]

[0127] At this point, the remaining rows of the sensing matrix A are transformed into cyclic shifts of c1(l). Calculate the discrete Fourier series of the modulation sequence and construct the sensing matrix A based on this series.

[0128]

[0129]

[0130] Step 8: Take the sampling sequence y[n] = {y1[k],...,y1[k]} obtained in Step 6 and convert it into a single sample. m [k]} is sent to the CTF module.

[0131] The processing steps for the CTF module are as follows:

[0132] Using the CTF module, the problem can be transformed into an MMV problem, thus enabling the estimation of the support set.

[0133]

[0134] Step 9: Perform eigenvalue decomposition on matrix Q from step 8 to obtain the right eigenvector matrix V.

[0135] Step 10: Solve the V=AU problem decomposed by the CTF module in Step 4 using the Orthogonal Matching Pursuit algorithm to obtain the joint support set.

[0136] The specific process of the orthogonal matching pursuit algorithm is as follows:

[0137] Step 11: Let the residual R0 = V, the sparse vector U = 0, and the support set The number of iterations l = 1;

[0138] Step 12: Find the best matching atom: where R l-1 Let a represent the residual after the last iteration. k Let <,> represent the k-th column of matrix A, and <,> represent the inner product.

[0139]

[0140] Step 13: Update the support set

[0141]

[0142] Step 14: Update residuals

[0143]

[0144]

[0145] Step 15: Verify if the iteration stopping condition is met: If the condition is met, i.e., l ≥ 2K or ||R l If ||2≤ε, then proceed to step 15; otherwise, let l=l+1 and return to step 12.

[0146] Step 16: Based on the joint support set obtained in Step 15 Restore the original signal.

[0147]

[0148] Step 17: Since the rate of z[n] is the same as the rate of y[n], it is necessary to process each sequence z... l Performing L-fold interpolation on [n], 1≤l≤L, increases the sequence rate to the sampling rate of the signal x(t), yielding...

[0149]

[0150] The interpolated sequence Through the ideal interpolation filter h lAfter [n], time-domain modulation and summation are performed to reconstruct the sampling sequence of the original signal.

[0151]

[0152] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the scope of the technology disclosed in the present invention, and such modifications or substitutions should all be covered within the scope of protection of the present invention.

Claims

1. A digital channelization sensing method based on MWC, characterized in that, include: Band-limited signals using Nyquist sampling rate The digital signal obtained by sampling is ; For digital signals Perform integer extraction to obtain a new data sequence. For data sequences Data is obtained by mixing. ; The digital signal Perform integer extraction to obtain a new data sequence. Specifically In each consecutive Take the first data from the data. Value, to obtain new data: In the formula, The extraction coefficient is a positive integer. To extract the phase; The data sequence Data is obtained by mixing. Specifically in, Represents a random modulation sequence The There are ... Each symbol has a random value of +1 or -1, and the equivalent symbol sequence periodic frequency is... , The sampling rate is Nyquist. Data Data is obtained through a multiphase filter bank with corresponding delays. , on data The signal is obtained by mixing. The polyphase filter bank is a prototype filter. The There are multiple phase components, and the normalized cutoff frequency of the prototype filter is 1 / L The data The signal is obtained by mixing. Specifically: Signal By interval D Divide and sum, then perform the following steps: M Point DFT obtains the signal ;in D and M satisfy ; The signal By interval D Divide and sum, then perform the following steps: M Point DFT obtains the signal Specifically: Signal The joint support set is obtained by sending it into the CTF module. According to the joint support set Restoring the original signal; the signal The joint support set is obtained by sending it into the CTF module. Specifically: Using the CTF module, the problem can be transformed into an MMV problem, thus enabling the estimation of the support set. : Where, vector The first in The element represents in Time system The sampled values ​​of each channel; any one of them satisfies matrix All can be called The framework, namely ,in It is a matrix eigenvalues The diagonal matrix formed It is composed of eigenvectors Arrangement dimensional matrix; if the equation There exists a unique joint sparse solution. And matrix If the RIP conditions are met, then It also has a unique joint sparse solution. , support set and support set Maintain consistency; utilize the CS reconstruction algorithm for the joint sparse matrix By making an estimate, the result can be obtained. joint support set ; The orthogonal matching pursuit algorithm is used to solve the CTF module decomposition. The problem is as follows: Let the residual sparse vectors Support set Number of iterations ; Finding the best matching atom: where This represents the residual after the last iteration. Representation matrix The List, Inner product Update support set Update residuals Until the iteration stops; According to the joint support set To restore the original signal, specifically: This method also includes: For each sequence conduct Double interpolation increases the sequence rate to the signal rate. The sampling rate can be used to obtain The interpolated sequence Through ideal interpolation filter Then, by performing time-domain modulation and summation, the sampling sequence of the original signal can be reconstructed. 。 2. A computer system, characterized in that... include: One or more processors, a computer-readable storage medium for storing one or more programs, wherein, when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method of claim 1.

Images