A Study On Singal Recognition And Detection Based On Compressive Sensing In Wireless Communication Systems

In recent years, to meet the challenge of spectrum sensing(SS) with ultra wide band and big data in cooperative and cognitive radio networks, the requirements of both the technology and signal processing capability urging the receiver to its ultimate load. Consequently, new technology of green communication with high efficiency has attracted more attention and got fast development in signal processing field.As a basic principle and method, Shannon-Nyquist Sampling theory has been regarded as the guidance in field of signal processing for a long period of time. The introduction of compressive sensing (CS) theory makes it possible for OFDM systems to achieve high efficiency and well performance in dealing with massive information issues. The technology of Compressive Sensing and Compressed Signal Processing (CSP) are being studied in all kinds of Applications.In this dissertation, to meet the requirements of signal recognition and detection in non-cooperative wireless communication systems, we work out the theoretical framework of signal recognition and detection with compressed samples, and build a test model to confirm the algorithm and methods we propose. By analyses of statistics parameters such as autocorrelation and high-order cyclic cumulants, we present both the process of signal processing and reconstruction algorithm under the method of compressed sampling. Moreover, simulations and algorithm verifications are conducted to get the conclusions and achievements.Inspired by the approach of CS reconstruction on account of spectrum analysis, we in the first place explore the sparse express of target signals, by studying on signal features of various dimensions, which can confront the lack of prior information under the condition of non-cooperative system. Then we successfully achieve the ideal reconstruction performance by sub-Nyquist samples using the sparse domain of cyclic autocorrelation function (CAF).Furthermore, to extract the statistical features of communication signal from compressed samples, such as cyclostationary, full-scale reconstruction is actually not necessary or somehow expensively. At the same time, direct reconstruction of cyclic autocorrelation function may not be practical due to the relatively high processing complexity. Therefore, we at last propose a new approach of cyclic feature indirectly recovery based on the reconstruction of autocorrelation sequence from sub-Nyquist samples. And it is proved that it can reduce both the computation complexity and memory consumption significantly, while the recovery performance maintains well in the same compressive ratio. Through theoretical analyses and simulations, we conducted to show and verify our statements and conclusions.