An Efficient MTD Processing Method Based on Conjugate Symmetric Structure

Abstract

The present invention relates to an efficient MTD processing method based on a conjugate symmetric structure, belonging to the field of radar signal processing, and includes: calculating pilot vectors corresponding to each Doppler frequency and calculating windowed equivalent filter coefficients; constructing a conjugate complex symmetric pilot vector matrix based on the pilot vectors by using frequency symmetry to obtain a conjugate complex symmetric FIR digital filter coefficient matrix; using the inverse of the interference covariance matrix R to match clutter to complete the correction of the FIR filter weight coefficients; using the characteristics of the improved Harvard structure of the DSP to perform a memory block pipelined architecture, and reconstructing the operation steps of the FIR filter according to the conjugate complex symmetric characteristics of the FIR filter to reduce the calculation amount, complete the filtering process of the filter bank, and obtain the final processing result. The present invention designs the FIR filter bank into conjugate complex symmetry, reconstructs the FIR filtering process, and combines ping-pong caching, an efficient instruction set, and pipeline optimization to achieve fast calculation results.

Description

Technical Field

[0001] The present invention relates to the technical field of radar signal processing, and particularly to an efficient MTD processing method based on a conjugate symmetric structure. Background Art

[0002] With the development of technology, the current radar combat scenarios are becoming increasingly complex, and the data rate required by radars is getting larger and larger, resulting in a sharp increase in the amount of data that the backend signal processing platform needs to process; currently, basically all radars are coherent radars. Due to the determination of the phase of the transmitted signal between periods and its reliable performance, it provides the possibility of extracting the Doppler characteristics of targets. It is precisely based on this coherent system of radars that when the radar operates in a complex environment, through MTD (Moving Target Detect), that is, moving target detection technology, clutter can be suppressed to achieve the detection of moving targets; MTD processing is to separate targets and clutter in different Doppler channels through a partially overlapping Doppler filter bank, thereby realizing the detection of moving targets by the radar, which is an important link in radar signal processing.

[0003] From the perspective of signal processing design, the implementation methods of MTD mainly include two types: fast Fourier transform (FFT) and FIR filter. In engineering, due to the wider theoretical basis of fast Fourier transform (FFT) on both software and hardware platforms, it is easier to implement using FFT. Compared with the FIR filter, the algorithm implementation complexity of FFT is lower, and the operation speed is faster. For the echoes of N periods, through M - point MTD processing using FFT, the output value of each point is the 1 / M - period frequency component of the M - point FFT processing, which is equivalent to an accumulation of N - period echoes at this frequency, thereby achieving coherent accumulation and obtaining the Doppler information of the target, realizing MTD processing. Since the number of points M to be processed in the FFT engineering implementation algorithm needs to be a power of 2, so M≥N. In most cases, the radar design cannot be exactly designed as a power of 2, while the FIR filter bank has no such requirement for the number of accumulation points. The FIR filter commonly designed and used is a finite - impulse - response (FIR) filter, which has good stability and ideal linear phase and is widely used. However, its main drawback is that the better the filtering performance, the higher the order required, and thus there are problems such as large time delay and many storage units required. For the traditional FIR filter, each operation data needs to be multiplied and accumulated one by one, so the time resources required are greater than those consumed by the fast Fourier transform (FFT). As the data volume increases, its impact on the processing speed is particularly obvious. As a dedicated digital signal processing device, DSP has powerful professional computing capabilities and is suitable for implementing more complex algorithm processing. However, its serial instructions limit its processing ability for high - speed signals. Under the limited processing time resources, it is not conducive to its real - time processing in more complex algorithms, thereby affecting the optimization and improvement of signal processing performance. Therefore, optimizing the MTD processing based on the FIR filter bank is an issue that needs to be considered currently. Summary of the Invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and provide an efficient MTD processing method based on a conjugate - symmetric structure, which solves the deficiencies existing in the prior art.

[0005] The purpose of the present invention is achieved through the following technical solutions: An efficient MTD processing method based on a conjugate - symmetric structure, the MTD processing method includes:

[0006] S1. Calculate the pilot vectors corresponding to each Doppler frequency and calculate the windowed equivalent filter coefficients;

[0007] S2. According to the windowed pilot vectors, use frequency symmetry to construct a conjugate - complex - symmetric windowed pilot vector matrix to obtain a conjugate - complex - symmetric FIR digital filter coefficient matrix;

[0008] S3. Use the inverse of the interference covariance matrix R to match clutter and complete the correction of the FIR filter weight coefficients;

[0009] S4. Utilize the improved Harvard architecture characteristics of the DSP to perform a memory block pipelined architecture, and reconstruct the operation steps of the FIR filter according to the conjugate complex symmetry characteristics of the FIR filter to reduce the computational complexity, complete the filtering process, and obtain the final processing result.

[0010] The specific content of step S1 is as follows:

[0011] S1-1. The pilot vector of the k-th Doppler channel is where, φ k (n) = 2πnf k T r , n = 0, 1,..., N - 1, N is the number of accumulated pulses, f k is the center frequency of the k-th filter, T r is the PRT repetition period, [·] T represents matrix transpose, and j represents the imaginary unit;

[0012] S1-2. Perform windowed equivalent filter coefficient calculation based on the pilot vector to obtain the windowed pilot vector where · represents matrix dot product, that is, truncate the frequency sampling response that meets the requirements with a window function, where win is the Doppler window coefficient.

[0013] The specific content of step S2 is as follows:

[0014] According to steps S1-1 and S1-2, obtain the windowed pilot vector Let the angular frequency component ω(k) = 2πf k T r , then φ k (n) = nω(k), design the angular frequency component into a symmetric characteristic, and then obtain the windowed angular frequency vector According to Euler's formula transformation, obtain the symmetric formula of the angular frequency vector where * represents complex conjugate, M is the number of Doppler channels, k = 0, 1,..., M - 1 is the filter number, when M is even, that is, conjugate complex even symmetry; when M is odd, conjugate complex odd symmetry.

[0015] The specific content of step S3 is as follows:

[0016] Adjust the weighted filter coefficients, use the clutter information to correct the weight coefficients, so as to more effectively suppress the clutter components. At this time, the weight coefficients are w kdenotes the weight coefficient of the k-th filter, R is the interference covariance matrix under the clutter-free condition, and R c is the clutter covariance matrix, is the noise power, and the interference covariance matrix I is the identity matrix. Matching the clutter means performing an inverse operation on R and multiplying it by the weight coefficient to form nulling suppression.

[0017] The specific content of step S4 includes the following:

[0018] S4-1: Utilize the Harvard architecture characteristics of the DSP to establish a ping-pong buffer in the high-speed data memory of the DSP. Segment the data matrix processed by MTD according to range gates and use DMA to move it in the background. If the MTD-processed data matrix is L×N, let the length of each data segment moved be L m , and the number of moves is m = |L / L m |. When L % L m > 0, m is incremented by 1, and in the last time L m = L % L m , where L is the range cell, N is the number of accumulated pulses, % is the remainder operation, and |·| is the integer operation;

[0019] S4-2: Utilize the high-efficiency instruction set of the DSP and the conjugate symmetry characteristics of the FIR filter to reconstruct the multiply-accumulate processing operation of the FIR on the DSP. The coefficient memory and data memory in the cache can be prefetched, and then complex multiply-accumulate pipelining operations can be performed; for even symmetry, the Doppler channels k and M-1-k are conjugate complex symmetric. When performing complex operations, the result of multiplying the imaginary part is accumulated with the opposite sign, and at the same time, the conjugate complex symmetric filtering result is calculated; similarly, the conjugate complex odd symmetry is also operated in this way, but it is calculated separately at the |M / 2| channel.

[0020] The present invention has the following advantages: An efficient MTD processing method based on a conjugate symmetric structure. By designing the FIR filter bank into conjugate complex symmetry, reconstructing the FIR filtering processing flow, and combining ping-pong caching, high-efficiency instruction sets, and pipeline optimization, the filtering results of conjugate symmetric channels can be calculated quickly; at the same time, it can be freely designed according to the length of the processor cache, and thus it can further enable multi-core DSP processors to perform multi-core parallel processing, reduce the operation time resources of the processor, and increase the real-time performance of MTD processing; it can be quickly designed for different engineering application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a schematic flow diagram of the present invention;

[0022] Figure 2 is a schematic diagram of the amplitude response curve of the FIR filter bank;

[0023] Figure 3Schematic diagram of the processing flow of the FIR filter bank of the present invention;

[0024] Figure 4 Schematic diagram of the time comparison between the original FIR processing and the conjugate-symmetric block matrix FIR processing. Specific implementation manner

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Usually, the components of the embodiments of the present application described and illustrated herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the protection scope of the present application claimed, but only represents the selected embodiments of the present application. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative efforts belong to the protection scope of the present application. The present invention will be further described below with reference to the accompanying drawings.

[0026] As Figure 1 shown, the present invention specifically relates to an efficient MTD processing method based on a conjugate-symmetric structure. The method first designs a windowed pilot vector, then performs conjugate complex symmetry extension on it to form a conjugate-symmetric FIR filter bank, then performs matching correction according to the clutter prior information, and then reconstructs the FIR filtering operation to achieve pre-reading of the weight coefficient cache and data cache, and performs conjugate complex pipelined multiply-accumulate calculations to obtain the filtering results of the conjugate-symmetric channels simultaneously. This process reduces the computational amount while also improving the FIR filtering operation speed by taking advantage of the access speed differences of different memories; specifically, it includes the following contents:

[0027] Step 1: As Figure 2 shown, assume that the number of radar MTD accumulation pulses N = 64, the number of Doppler channels M = 64, the PRT repetition period T r = 120 μs, and the passband coverage width of a single Doppler channel is f = f r / M = 130.2083 Hz. It can be obtained that Then where n = 0, 1,..., N - 1, and the pilot vector is obtained. The pilot vectors of each Doppler channel are a(f 0 ),..., a(f M-1 ). Generate a window coefficient win of length M, such as a 60 dB Chebyshev window coefficient, and window the pilot vector to obtain the windowed pilot vector

[0028] Step 2: Let the angular frequency component Construct ω(k) = 2π - ω(M - 1 - k), and transform according to Euler's formula: e jω(k) =(e j(-ω(k)) ) * =(e j(2π-ω(M-1-k)) ) * , to obtain the angular frequency vector symmetry formula M is the number of Doppler channels, and k = 0, 1,..., M - 1 is the filter serial number (Doppler channel number); when M is even, it is conjugate complex even symmetry, that is and and and are conjugate complex symmetric. When M is odd, it is conjugate complex odd symmetry, that is and and and are conjugate complex symmetric, where |·| is rounding.

[0029] Step 3: Considering the suppression of ground clutter and noise, adjust the weight coefficients of the filter accordingly, that is, use the prior information of clutter to correct the weight coefficients, so as to more effectively suppress the clutter component; Gaussian white noise covariance matrix Ground clutter covariance matrix R c = E{CC H}, where C represents the clutter vector, E{·} represents the mathematical expectation, [·] H represents the conjugate transpose. Let the ground clutter spectrum deviation be According to radar principles, for a small mountain with trees, its speed standard deviation is generally not greater than 0.32 m / s. It can be assumed that the expected clutter speed standard deviation σ v = 0.15 m / s, and the transmit frequency is 16.25 GHz, then λ is the transmit wavelength; the clutter covariance matrix is obtained from the ground clutter autocorrelation function where p, q ∈ [1, 2,..., N], N is the number of pulses, is the clutter power, usually designed with the clutter-to-noise ratio CNR; let the white noise power The expected clutter-to-noise ratio CNR = 40 dB, then Further obtain the interference covariance matrix where I is the identity matrix; according to the weight coefficient formula, after inverting the interference covariance matrix R, correct the weight coefficient , that is, the weight coefficient of the kth filter Furthermore, obtain the filter coefficient group that forms a null suppression for the expected clutter.

[0030] Step 4: As Figure 3As shown in the figure, query the official DSP data and general tests. The DSP memory characteristics are shown in Table 1. Utilize the Harvard structure characteristics of the DSP to establish a ping-pong cache in the high-speed DSP memory, and segment the data matrix processed by MTD according to range gates and use DMA to move it in the background; this will facilitate the pipelining design of the FIR filtering process in MTD processing. Assume that the MTD processed data matrix is L×N, where L is the range cell and N is the number of accumulated pulses; let L = 2000 and N = 64; the length of the data segment moved at one time is L m = 40 range cells, and the number of moves m = |L / L m | = 50. Since L % L m = 0, so m is 50 times.

[0031] Table 1. DSP Memory Characteristics Table

[0032]

[0033] The direct form FIR filtering process is also called the convolutional or transversal FIR filtering process. The input data matrix is x, which is N rows and L columns, and the output data matrix is y, which is M rows and L columns. Then the filtering operation form for the k-th channel of the l-th range cell is: In the even symmetry, because the filter coefficients of the k-th channel and the symmetric channel M-1-k are conjugate complex symmetric, that is Therefore, the filtering operation for the conjugate even symmetric channel M-1-k of the l-th range cell is: where w knr 、w kni are the real and imaginary parts of the n-th sequence weight coefficient w kn of the k-th channel, x lnr 、x lni are the real and imaginary parts of the n-th row data x ln of the l-th range cell, where n = 0, 1,..., N-1, l = 0, 1,..., L-1, k = 0, 1,..., M-1, and j represents the imaginary unit; thus, using the efficient DSP instructions, reconstruct the FIR multiply-accumulate processing operation steps, and prefetch the coefficient memory and data memory in the cache, and perform the accumulation pipelining operation during the complex multiplication cycle of the coefficients and data; in the even symmetric FIR coefficient matrix, due to the conjugate complex symmetry of the Doppler channel k and the channel M-1-k, the complex operation accumulation operation of the channel k, while the channel M-1-k takes the negative of the imaginary part multiplication result of the channel k complex operation and accumulates it, and the filtering results of the channel k and the conjugate channel M-1-k can be calculated simultaneously; similarly, for the odd symmetric FIR coefficient matrix, except for the |M / 2| channel where the filtering result is calculated separately, the other channels are processed in the same way as the even symmetry to obtain the filtering result.

[0034] Such as Figure 4As shown, the method of the present invention provides a comparison of the processing time of the FIR filtering of the original algorithm and the FIR filtering of the optimized algorithm. The data matrix L = 2000 is the processing range cell, and the processing time for different numbers of accumulation pulses is compared. It can be clearly obtained that under the conditions of common accumulation pulse numbers in radar, the FIR filtering time of the optimized algorithm is significantly better than that of the original FIR filtering process, and at the same time is better than or close to the FFT processing time, demonstrating the superiority of the efficient MTD processing method with conjugate symmetric structure. In view of the fact that the method of the present invention is at the key position of MTD processing in traditional radar signal processing, optimizing the FIR processing algorithm and process to save its processing time has good engineering application value.

[0035] The above are only the preferred embodiments of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and improvements, and can be changed within the scope of the concept described herein through the above teachings or the technology or knowledge in related fields. And any changes and modifications made by those skilled in the art without departing from the spirit and scope of the present invention shall fall within the protection scope of the appended claims of the present invention.

Claims

1. An efficient MTD processing method based on a conjugated symmetric structure, characterized in that: The MTD treatment method comprises: S1, calculate the pilot vector corresponding to each Doppler frequency, and calculate the windowed equivalent filter coefficient; S2. According to the windowed pilot vector, a conjugate complex symmetric windowed pilot vector matrix is ​​constructed by utilizing frequency symmetry to obtain a conjugate complex symmetric FIR digital filter coefficient matrix; S3, using the interference covariance matrix R to find the inverse matching clutter, and complete the correction of the FIR filter weight coefficient; S4, using the Harvard structure characteristics improved by DSP to perform memory block pipeline architecture, and reconstructing the FIR filter operation steps according to the conjugate complex symmetry characteristics of the FIR filter, reducing the amount of calculation, completing the filtering process, and obtaining the final processing result; The S1 step specifically includes the following contents: S1-1, the pilot vector of the kth Doppler channel is ,in, , n=0,1,...,N-1, N is the number of accumulated pulses, f k For the k The center frequency of the filter, T r is the PRT repetition period, represents the matrix transpose, j represents an imaginary unit; S1-2, calculate the windowed equivalent filter coefficient based on the pilot vector to obtain the windowed pilot vector ,in represents matrix dot product, that is, truncation of the frequency sampling response window function that meets the requirements, where win is the Doppler window coefficient; The S2 step specifically includes the following contents: According to steps S1-1 and S1-2, the windowed pilot vector is obtained. ; Assume the angular frequency component ,but , the angular frequency components are designed to be symmetrical, and then the windowed angular frequency vector is obtained According to the Euler formula, we can get the symmetric formula of the angular frequency vector ,in * represents complex conjugate, M is the number of Doppler channels, k=0,1,...,M-1 is the filter number, when M is an even number, the conjugate complex number is even symmetric; when M is an odd number, the conjugate complex number is odd symmetric.

2. The method of claim 1, wherein: The S3 step specifically includes the following contents: The weighted filter coefficients are adjusted and the weight coefficients are corrected using the clutter information to more effectively suppress the clutter component. At this time, the weight coefficient is , w k Indicates k The weight coefficients of the filters, R is the interference covariance matrix under the condition of no moving clutter, R c is the clutter covariance matrix, is the noise power, and the interference covariance matrix is ​​obtained , I is the unit matrix, and matching clutter is to perform the inverse operation on R and dot multiplication with the weight coefficient to form null suppression.

3. The efficient MTD processing method based on conjugated symmetric structure according to claim 1, characterized in that: The S4 step specifically includes the following contents: S4-1. Using the Harvard structure characteristics of DSP, a ping-pong cache is established in the DSP high-speed data memory. The data matrix processed by MTD is segmented according to the range gate and moved using the DMA background. If the data matrix processed by MTD is L×N, the length of each data segment moved is L. m , the number of moves is , when L%L m >0, m is incremented by 1, and the last L m =L%L m , where L is the distance unit, N is the number of accumulated pulses, and % is the remainder. To round up; S4-2, using the DSP efficient instruction set and the conjugate symmetry characteristics of the FIR filter to reconstruct the multiplication and accumulation processing operation of the FIR on the DSP, pre-read the coefficient memory and data memory of the cache, and then perform complex multiplication and accumulation pipeline operations; for even symmetry, the Doppler channel k Symmetrical to the conjugate complex of channel M-1-k, the imaginary part multiplication result is negated and accumulated during complex number operation, and the conjugate complex symmetric filtering result is calculated at the same time; The same is true for the conjugate complex odd symmetry, but in The channels are calculated separately.

Images