Automatic Recognition Of Communication Signal Modulations

Modulation recognition of communication signals is an important problem in signal processing. The objective is to decide the modulation type and estimate the modulation parameters of the intercepted signals without any priori knowledge about the signal contents, and to provide reference for farther signal processing. At first, the dissertation studies feature extraction based on wavelet transform, carrier frequency estimation and symbol rate estimation, etc. It lays the groundwork for future signal identification. For the modulation recognition, in consideration of the differences between analogue and digital modulated signals, the dissertation studies the two kinds of signals respectively:a coarse classification algorithm using wavelet transform is described, and an improved method is proposed to extract the state of spectrum peaks more directly, then on the basis of the method, another coarse classification using nonlinear transform and wavelet transform is presented; the specific modulation type of analogue and digital modulated signals is identified from time domain and transformation domain, which utilizes the classical method and novel way based on wavelet transform and spectrum feature analysis respectively; the associative recognition of all signals shows the independence and expansibility of the algorithm that is combined by coarse classification and specific classification. In the end, ANN and SVM are applied to design the classifiers of modulation recognition, and the good application potential of SVM used in modulation classification is revealed.In the dissertation,15 kinds of signals are considered, such as AM, DSB, LSB, USB, FM,2ASK,4ASK,8ASK,2FSK,4FSK,2PSK,4PSK,8PSK,16QAM and 64QAM. The simulation results indicate that the present scheme is feasible and effective, and the recognition performance is good.