Research On Modulation Mode Recognition Algorithm Of Digtal Communication Signals

With the rapid development of communication technology, communication system and the modulation mode are increasingly diverse, form the coexistence of a variety of communication systems, and have brought great difficulties for communication Internet between multi-institutional. In order to receive the multiple modulated signals at the same receiver platform, modulation mode recognition technology between signal detection and signal demodulation, has been widely studied.In recent decades, many scholars at home and abroad research on it and acquire a lot of results. But there are still serious deficiencies, especially in the case of no a priori knowledge, algorithms overall have difficulty extracting stable feature parameters and be low recognition in low SNR conditions; Meanwhile, communications modulation automatic identification is still not a universal recognition algorithm, and most of the research work are for certain kinds of signals, but other signals algorithm failure; so recognition rate and the range are still the bottleneck of digital communications modulation mode automatic identification technology development, and have become a top priority of the current study. According to these problems, the paper researches the identification of digital communication signal modulation, started a discussion around the identification of common digital signal. The main contents are summarized as follows:1) The feature extraction methods are analyzed theoretically and simulated among modulation recognition. It can be summed up the strengths and weaknesses of the instantaneous features, fractal theory and wavelet theory used in feature extraction for specific digital communication signals. It improves in high order cumulants characteristics of modulation recognition algorithm, namely combined with high order cumulants and wavelet extraction with excellent performance characteristic parameters. The improved algorithm reduces the computational complexity, and improves the recognition rate.2) The paper studies the classification methods of the modulation recognition. The decision tree and the support vector machine of statistical learning theory are analyzed in statistical pattern recognition methods. So choose the support vector machine as classifier. It makes improvements on the basis of one-to-one, one-to-many and acyclic graph multiple signal classification methods, and based on binary tree support vector machine modulation recognition algorithm is proposed, which improves the recognition performance in further.3) According to the actual communications environment, the improved algorithm based on spectral correlation function and some new classification features are proposed in multipath channel. Anti-multipath interference characteristics of the new identification features are analyzed theoretically and verified experimentally. Through the use of the frequency smoothing method, analyze the characteristic spectrum in cyclic frequency axis of digital communication signals, extract characteristic surfaces and projection surfaces of the cyclic spectrum of digitally modulated signals, the quadratic cyclic spectrum and the fourth power cyclic spectrum, use the correlation coefficient of these as characteristic parameters, complete the multi-carrier and single carrier signals recognition, expand the scope of the recognition and achieve better recognition results.