The Research Of Modulation Recognition Method Under The Lower SNR

Modulation identification for communication signals is a still important problem in the intercepted signal processing, it is required to identify the modulation format and modulation parameters in the complicated signal environment with noise, and to provide reference for farther analysis and processing. Modulation format is one of the most important characteristics used to distinguish communication signals. After analyzing the received signal, the objection of modulation identification is to decide the modulation format and estimate the modulation parameters of the communication signal without any prior knowledge about the signal information content. With the development of communication technology, the spatial signals are more and more complicated and dense. It results in that output of the receiver doesn't contains only one signal, which makes the modulation identification more difficult, and as results, there comes more demands for the research of modulation identification of communication signals.It could be seen from recently published articles in home and abroad, more and more attentions are being paid in the research of automatic modulation identification theory of communication signals, especially in the modulation identification under low signal to noise ratio. This dissertation performed the following work based on the research before:(1) We studied the modulation identification problem of analogue modulated signals. In condition of the low SNR and without prior information, we successfully identified AM,DSB,USB,LSB and FM signals by using signal high order envelope characteristic and spectrum symmetry parameter. The two parameters are simple and have good distinguish characteristic. And then, we designed identifying process.(2) We studied the modulation identification problem of digital modulated signals. First, we introduced the signal is preprocessed by using raised cosine filter, and then abstracted five characteristic parameter, identified the digital signal by using decision tree classifier and neural network classifier. The decision tree classifier needs threshold of decision, however, the choosing of threshold of decision is fussy, and the human factor very large, so the recognition performance depend on the threshold of decision severely. The neural network classifier adopt multilayer combinatorial neural network classifier, comparing with traditional methods, the network model and training algorithm designed in this paper is improved much in convergence speed, training time and recognition ratio.(3) We studied two kinds of modulation identification problem of digital modulation signals and analog modulation. The signals can be divided in two parts, analog signals and digital signals by using a way of abstracting characteristic parameter after nonlinear transformation, this way is simple, and has smaller operation and higher recognition rate. This is a base for further modulation recognition.