A Research Of Modulation Recognition Based On Support Vector Machines

With the rapid development of communication technology, satellite communications play a vital role in both civilian and military, and modulation of communication signals in communications technology research is to identify critical. In view of the excellent performance demonstrated of the support vector machine method used in the area of modulation recognition, this paper study and simulation analysis for commonly digital modulation identification algorithm used in satellite communications based on support vector machine classification method.This article firstly analyzes the characteristics of variety of commonly modulation in satellite communication, studies on the classification of modulation recognition, analyzes and compares the decision tree and the method of support vector machine part of the statistical learning theory in statistical pattern recognition method , and chooses the method of support vector machine as the classifier; Secondly, analyzes the commonly feature extraction methods, selects the higher order cumulants as feature extraction methods in the process of identification in accordance with the characteristics of modulation signal used in satellite communications; then, extracts the characteristics of the signal modulation based on the method of higher order cumulants and puts forward the improved feature extraction program for the modulation signal which are too difficult to classify, that is combined the mathematical methods such as the Fourier transform, the differential transform, the difference of phase, with the higher order cumulants and extracts the characteristics of excellent performance; finally, through the study of relevance theory about the support vector machine, simulates and analysis the identify processes of the commonly satellite communication modulation method used the multi-category classification method of tree-based as a classifier, simulates and studies on the related issues of support vector machines , including the performance comparison of constructor of a variety of multi-category classifier and the impact of different kernel function used in identify modulation. The research shows the combination of that support vector machines with higher-order cumulants and improved feature extraction programs gets a higher recognition rate in the identify process of commonly digital modulation methods used in the satellite communications , has a good performance and prospects for broader research.