Research On Compression And Reconstruction Method Of Communication Signal Features

Wireless communication signal detection and recognition technology as an important part of cognitive communication and spectrum sensing technology,has been concerned and researched diffusely.In modern communication systems,signal detection methods based on feature extraction have been widely used.However,due to the limitation of Shannon/Nyquist sampling theory,spectrum sensing system needs to deal with huge amounts of data and put forward very high requirements for physical equipment.By introducing compressive sensing theory,the signal can be sampled at a low bit rate to effectively alleviate the pressure of processing data for system.In this paper,compressive sensing theory is used to study the compression and reconstruction of communication signals.Firstly,by analyzing the theoretical construction model of signal statistical characteristics,the structural decomposition method of partial signal statistical characteristics is proposed.The decomposed sparse representation is taken as the reconstruction target to realize the optimization of the reconstruction process.In this method,the resolution of the structural decomposition matrix depends on the location information of the feature,and the decomposition basis is simple,which reduces the complexity of the sensor matrix and the memory space consumed by the sensor matrix.At the same time,the reconstruction process takes the substructure as the iteration unit to significantly reduce the number of iterations.Furthermore,the k-model sparse theory is introduced to further constrain the target representation space,which reduces the search cost of the compression sampling pursuit reconstruction algorithm,and achieves the goal of optimization reconstruction process.In addition,the eigenvalue decomposition and reconstruction method of single signal feature is proposed by studying the eigenvalue decomposition property.This method not only guarantees the high sparsity of the decomposition results,but also has the advantage of reconstructing most of the original features with less iterations.Based on the above method,singular value decomposition is further introduced,and a common sparse basis extraction method for similar features of multiple signals is proposed,which supports the use of one sparse basis to accomplish the similar feature reconstruction of multiple signals at the receiver.Finally,based on the proposed two methods,we give a compression reconstruction algorithm and process of autocorrelation and fourth order time-varying moments.The feasibility of the algorithm is verified by Matlab simulation,and the performance of the algorithm is analyzed from the sparsity of the decomposition results,the complexity of reconstruction,and the reconstruction error.