Research On Modulation Recognition Of Communication Signals Based On Noise Robustness

With the rapid development of communication technology, the communication signal modulation mode identification technology based on pattern recognition is an important research topic in the field of software radio. After years of research and development a lot of research has been made, but the research about large dynamic signal-to-noise ratio under the environment of communication signal modulation recognition has been unable to obtain good results, and the use of manual feature selection methods will increase much work.In this paper, we study the method of modulation recognition of communication signals with good classification ability under the condition that the SNR is dynamic and fast changing.Firstly the involved 10 kinds of modulation signal that include MASK, MFSK, MPSK and 16 QAM signals have been analyzed and explained. After that we extracted from the instantaneous information characteri stics of communication signals, statistic features and frequency domain features, and the use of signal features extracted group into the original feature set..Secondly, through the analysis and research of deep belief networks(deep belief network, the DBN), the DBN network is applied to extract the original feature sets to get the new feature set and the new feature set has the noise robustness. This approach not only reduces the dimensionality of the feature, but also increases the computational speed. The characteristic number of the feature set is reduced to 4, and the 4 kinds of features are not sensitive to the change of SNR.Then we use the SVM algorithm and the depth of learning neural network to classify the characteristics of the filter, and compare with the results of other literatures. The experimental results show that the classification accuracy of the SVM classifier is more than 99% in the 0d B-20 d B signal noise ratio environment.Finally, this essay studies the method of a kind of modulation recognition of communication signals with good classification ability, which has a good classification ability in the environment in which SNR is dynamic and fast changing and verifies its effectiveness.