Parameter Estimation Of Multi-band Signal Based On Compressive Sensing And Cyclic Spectrum

Multi-band sparse signals are widely used in passive radar design,ultra-wideband communications,and spectrum sensing in cognitive radios.Due to the unique sparsity of multi-band signals in the frequency domain,the theory and implementation based on compressed sensing will bring far-reaching influence to these application.In these fields,signal receivers usually receive one or more unknown signals at the same time.These signals differ in carrier frequency,symbol transmission rate,bandwidth,and modulation method.The receiver needs to perform fast recognition and response for signal parameters,through the modulation method identification to determine the signal modulation mode,and then perform other processing tasks.Therefore,identification and detection of various parameters of unknown signals becomes a crucial technology in the perception and acquisition of unknown information.In this paper,we mainly use the cyclostationary characteristics of multi-band signals to extract the parameter information of the signal.Since the estimation of the carrier frequency and bandwidth of multi-band signals requires extremely high sampling frequency,the theory of compressed sensing is introduced to reduce the ADC's requirements for performance.Combining compressed sensing theory with cyclic spectrum theory,the parameters of carrier,carrier frequency,symbol rate and modulation method of multi-band communication modulated signals are identified and estimated.The main research contents are signals parameter identification method based on compressed sensing and cyclic spectrum theory.The research scheme starts from the aspect of compressed sensing,and combines the detection and recognition of communication signals in the subject.First,the sparse characteristics of the statistical features of the communication signal and the sparse decomposition method are used to start the process.Because the signal is divided into blocks in the process of calculating the cyclic spectrum of the signal,so the sparsity and location of sparsity of the signal within each block are the same,and the SOMP algorithm can be used to reconstruct the signal's complex demodulation.According to t the cyclic spectrum of the complex demodulation of the signal estimate signal carrier frequency,modulation and symbol rate jointly.Since the multi-band signal is composed of a plurality of sub-band signals,the bandwidth,the carrier frequency,and the modulation method used by each sub-band signal are different,and the receiver has no knowledge of parameter of the received signal.In the process of estimating the signal cycle spectrum,since the complex demodulation of the signal contains part of the frequency spectrum information of the received signal,this part of the information can be extracted and analyzed first in the analysis of the cyclic spectrum of the compressed sensing,and the signal parameters such as carrier frequency,signal bandwidth,and number of signals were estimated.Each sub-band signal has a different modulation method.Because the communication modulation signal does not have the characteristic of the modulation method in the frequency domain,but the different modulation signals show uniqueness in the double-frequency plane of the cyclic spectrum.Therefore,the neural network pattern recognition is applied to the cycle characteristics and then the modulation method corresponding to the cyclic spectrum was obtained.Finally,this paper simulates the proposed cyclic spectrum estimation algorithm and parameter identification algorithm based on multi-band signal compressive sensing,and compares it with some existing algorithms,and analyzes in detail the performance comparison of the algorithm when estimating the signal cycle spectrum.The computational complexity of the algorithm is calculated.Then through the recovered signal cycle spectrum,the parameters of communication modulation signals such as BPSK,MPSK,QAM and other signals are identified and the robustness of the algorithm is analyzed and studied.The final result shows that the proposed algorithm has greatly improved recognition performance and computational complexity.