Research On The Sampling And Signal Processing Technology Of Wideband Signals Based On Compressive Sensing

With the rapid development of wireless communication technology, the electromagnetic environments become increasingly complex. In the electromagnetic spectrum monitoring, radio spectrum sensing, communication reconnaissance, and other non-cooperative receiving applications, not only the frequcncy bands for processing become increasingly wide, but aslo the signal types become more and more abundant, and the dynamic range of signal becomes increasingly large. As a result, the receiver must have the receiving and processing ability of wideband, large dynamic range signals. However, because of the sampling capability and the dynamic range limited by the analog to digital converter(ADC) device, the wideband signal acquisition and processing system based on Nyquist theory has faced serious technical challenges.For sparse signals or compressible signals, compressive sensing theory sample the signals use a frequency far below the Nyquist sampling rate. So reduce the sampling and dynamic range requirement for analog to digital converter in receiving end, and relieve the pressure on the digital end of a large amount of data processing, so to provide new ideas for solving the wideband signal acquisition and processing problems.This paper applies compressive sensing in broadband signal acquisition and processing with wide bandwidth and large dynamic. Sparse signal reconstruction under the condition of sensing matrix uncertainty, quantization spectrum and spurious-free dynamic range analysis of compressive sensing, strong interference suppression in compressed domain, signal detection and modulation recognition based on compressive sensing, implement of electromagnetic spectrum monitoring based on AIC have been mainly researched. The main work and innovations in the article are as follows:1. For the implementation issues of compressive sensing in wideband signal acquisition, considering the problems that the measurement matrix perturbation and sparse matrix mismatch leads to poor performance of traditional signal reconstruction algorithm, a compressed measurement model of sensing matrix uncertainty is derived, and proposed joint regularization reconstruction algorithm based on the error limits information. Theoretical analysis and simulation results show that compared with the standard compressed measure model and reconstruction algorithm, the proposed algorithm can effectively enhance the performance of reconstruction. Standard compressive sensing reconstruction algorithms are based on ideal mathematical model, however when exist measurement matrix perturbation and sparse matrix mismatch, the reconstruction performance dill degrade because of the measurement error. Based on the analysis of the cause of sensing matrix uncertainty and take advantage of the error limits of perception matrix and measurement vector, The sparse solving problems of perception matrix uncertainty was converted a L1 and L2 norm joint regularization convex optimization problem, constraint perception matrix uncertainty by L2 norm under sparse constraint. Complete sparse signal reconstruction at the same time ensuring the stability of solving at the expense of the signal sparsity.2. For the SFDR performance of wideband signal receiver, First, it theoretically deduced the theoretical performance bound of SFDR in the condition of sinusoidal signal input, and analyzed the influence of quantization bits, Gaussian noise and sampling rate on the performance, and the relevant conclusions are arrived. And then the quantization noise spectrum of compressed measurement and the SFDR performance have also analyzed in theory and deduced in mathematics, So that relevant conclusions are arrived. Compared with traditional ADC sampling, the SFDR performance of compressive sampling is influenced less by noise and the non-linearity of ADC, and enhance the SFDR performance of wideband signal sampling by lower the sampling rate and discard a little saturation value. The test method of ADC SFDR performance cannot arrived accurate results because of the influence of device, on the basis of the analysis in the ADC quantization noise spectrum of uniform quantization, it is theoretically derivate the performance of SFDR which incented by mono sine signal and concluded that SFDR performance improved when no noise there and variance of additive Gaussian noise influences the SFDR performance. Combining with Fourier analysis method, it analysis the quantization noise spectrum distribution in the condition of different sampling rate, and the relationship between the performance of SFDR and the sampling rate, and deduced that the SFDR performance is best when the sampling rate is prime relation with sine excitation signal frequency and the SFDR performance rise by line when integer times sampling. Further analyzed the quantization noise spectrum of compression measurement, and it concluded that because of the role of the random measurement matrix, quantization noise spectrum of compressive sensing is the white noise spectrum which is nothing to do with the input signal form. Finally, it analyzes the ADC circuit nonlinear influence and quantization bits on the SFDR performance of compressive sensing, from the aspects of decreasing of sampling rate and the fairness of measurements to illuminate the feasibility that the compressive sensing can solve the large dynamic problem in wideband signal acquisition.3. For strong interference suppression problems in broadband signal acquisition and processing, this paper proposes a strong interference suppression algorithm which is based on the minimum output energy criterion in compressed domain. The theoretical analysis and simulation results show that in the case of interference signal support set information is unknown, this algorithm can effectively restrain the strong interference effect on the properties of the target signal reconstruction. Under compressive sensing, it mainly uses the orthogonal projection algorithm and subspace oblique projection algorithm to suppress interference, but these two kinds of algorithms need to be on the premise of prior knowledge of jamming signal support set, which often can not be met in the non-partner broadband signal acquisition and processing. Therefore, this paper proposes an interference suppression algorithm which no need for support set a priori knowledge. This algorithm,making the minimizing output energy for the expectation projection value of each column of the sensing matrix as the principle, designs corresponding projection filter to projection filtering compressed measurements. Further it sets the threshold of projection value for suppress interference signals, meanwhile, preserving all the information in the target signal for subsequent processing.4. For the problem of sparse signal detection and identification of the signal modulation which in the condition of low signal-to-noise, this paper, introducing the ideas of cyclic spectrum analysis to the compressive sensing framework and based on the block sparse feature for cyclic frequency section, proposes a signal detection algorithm which based on compressed cycle energy spectrum feature. It can reduce the computational complexity effectively, and ensure the detection performance as well. On this basis, it designs a cycle spectrum feature extraction method for compressed domain which based on binary iteration, and combined with the binary tree classifier meanwhile, to achieve the identification of the modulation signal. First, it analyzes the detection performance limitations of the existing compressed detection and subspace detection algorithm of compressive sensing framework under the condition of low SNR. And then based on the fact that most of the modulation signal is with smooth circulation features and Gaussian white noise only emergent in the zero cyclic frequency, this paper introduces the cyclic spectrum analysis into the compressive sensing framework and proposes the sparse signal detection algorithm in the basic of compressed domain cyclic spectrum energy characteristics. This algorithm is different from the existing compressed cycle spectrum signal detection algorithm. It does not need to perfect reconstruction for the compressed cycle spectrum and makes full use of block sparse feature of signal in cyclic frequency section, so that the required number of compressed measurement greatly reduces. The simulation results show that this algorithm can effectively realize signal detection in low SNR conditions. Finally, on the basis of sparse block compressed cycle spectrum model, it gives an extraction method for cyclic spectrum feature which based on binary iteration and also combines with binary tree classifier for five kinds of common signal modulation(BPSK, FSK, 2ASK, 16 QAM, MSK) recognition.5. According to the requirements of project "Research on the Key Technologies of Electromagnetic Spectrum Monitoring Sensor Network", an electromagnetic spectrum monitoring program and principle of verification of hardware platform based on AIC is designed, and a method of correction based on adaptive filtering is proposed to improve the reconstruction performance when the filter impulse is non-ideal. In order to reduce the probability of undetected instantaneous burst signal, this program sampling the electromagnetic spectrum signal in wideband based on AIC compression scheme, and using the compressed measuring value to implement the signal detection, modulation recognition,etc of signal processing in compressed domain. Design and implement the verification hardware platform of the program, using RD structure an analog end, taking into account system scalability and flexibility using GPU + FPGA + ARM structure in the digital end. Finally to overcome the problems of non-ideal effects of the filter impulse response on signal reconstruction performance, design adaptive filter correction algorithm to estimate the non-ideal impulse response, and without changing the structure of the compressed measuring.