The Research On Recognition Of Communication Signal Based On Feature Extraction

Modulation recognition is a key technology in non-cooperation communication,which plays an important role in military and civilian fie lds.Such as in non-cooperative communication, after completing signal detection and some parameter ’s estimation, in order to obtain specific content information of the signal, we must first determine the mode of the modulation signals by modulation recognition, which also provides a basis for other following communication tasks. For the electronic warfare in the field of military,before listening or attacking communications of the enemy,we must first identify the mode of communication signals.In civilian area,recognizing of signal and monitoring and measuring of radio spectrum can not deviate the technology of modulation recognition.Therefore, either from the point of technology or application, the technology of modulation recognition research on communication signals is very necessary and important.In recent decades, scholars have gotten a lot of achievements in modulation recognition research, many of the new method is proposed. The two kinds of modulation recognition techniques for communication signal,the recognition based on likelihood ratio decision and the recognition based on statistical pattern,are introduced.The principles and characteri stics of the recognition based on statistical pattern are reviewed in detail. The advantages and disadvantages of the signal features in time domain and frequency domain classification performance are analyzed and summarized. And the extraction methods of higher-order cumulant features and cyclostationarity are discussed particularly. Then, this paper introduces the definition and properties of higher order cumulant, gives the signal MASK(2 ask, 4 ask, ask 8), MPSK(2 PSK, 4 PSK, 8 PSK), MFSK(2 FSK, 4 F SK, 8 FSK) the higher order cumulant theoretical value. Recognition classifier was designed according to extract the higher-order cumulant features difference, and the experimental results show that the recognition algorithm has good recognition effect after the MATLAB simulation. On the one hand, higher order cumulants feature recognition method from the perspective of the time domain to begin, in order to achieve complementary recognition performance, it can be further analysised by transforming into the frequency domain.On the other hand taking into account these signals can not be fully identified by the classification method b ased on higher-order cumulant, the identification method based on cyclic spectral characteristics can be further explored.So the MASK, MPSK, the cycle of MFSK signal spectrum characteristics are analyzed, and the four types of cyclic spectrum characteristic parameters for classification are showed.