Research On The Algorithm Of Features Extraction And Modulation Recognition Of Communication Signals

Modulation recognition of communication signals is one of the core technologies in the field of Software Radio(SR) and is also a hot issue in the signal processing field in recent years. With rapid development of communication technology, the modulation manner of communication signals have become more and more complicated and diversified. It results in that the routine methods and theories of recognition can hardly satisfy practical requirements and can’t effectively recognize communication signals. So the strict demand has been presented for further study on recognition of communication signals.At present, analog modulation signals are very limitted, and the methods of parameters extraction and modulation recognition have developed very well. And there are all kinds of digital modulation signals whose modulation styles are complex. With the development of communication technology, more and more digital modulation types will emerge in succession. Therefore, the study of features extraction and modulation recognition of digital modulation signals is the focus in academic researches. So this paper mainly focuses on seven kinds of digital modulation signals, namely 2ASK,4ASK,2FSK,4FSK,2PSK,4PSK and 16QAM.The general steps of signal modulation recognition include signal preprocessing, feature extraction and classification.In the preprocessing stage of modulation recognition, polyphase filter is used to extract the inphase component and the quadrature components, which can effectively reduce the sampling rate of the signals.In the features extraction stage of modulation recognition, six instantaneous feature parameters of seven kinds of digital modulation signals are extracted, among which the Roa is an improvement and the Rσp is an analogy of the Rσa.The simulation experiments show that the six instantaneous feature parameters are very good classification performances.In the design of classification stage of modulation recognition, a Particle Swarm Optimization algorithm with Neighbor information (NPSO) is proposed. In the NPSO algorithm, the search strategy of the basic PSO algorithm is preserved, at the same time, the local search of the neighbor particles is considered. That further improves the global optimization performance of the algorithm. By the NPSO algorithm to optimize the weights and thresholds of BP neural network, the NPSO-BP classification is designed. The simulation experiments show that the recognition rates of seven kinds of signals can be achieved more than 86% when the signal-to-noise ratio(SNR) is low to OdB, which proves that the classification can effectively improve the recognition performance of modulation signals.