Analysis And Identification Of Digital Modulation Signals Based On Wavelet And Fractional Derivative

The basic tasks of radio signals modulation identification are analysis, decision and classification of modulation type for intercepted unknown signals. So, the signal features should be specified in advance and their value ranges related to modulation types are determined. Then, the features of the signals are measured, and the measurement results are used to identify the modulation types. Thus, the basic important step of modulation identification is how to specify and extract signal features.There are three basic types of digitally modulated signal: MASK, MPSK and MFSK. Modulation identification includes inter-class classification and intra-class classification. About inter-class classification, the variances of amplitude and frequency are used to distinguish ASK, FSK, and zero-crossing analysis is used to tell FSK from PSK. With respect to intra-class classification, the variance of amplitude can be used to identify MASK, spectrum analysis is applied to identify MFSK, and MPSK identification is realized by checking the phase variation when the symbol changes.Wavelet transform can be reviewed as a bank of constant-Q filters according to signal processing theory, thus it is suitable for transient detection and digital signal modulation identification. The fractional order reflects the smoothness of signals, and it also completely specifies the time-frequency localization feature. Therefore, fractional derivative is good at describing singularities of signals, and can be applied to analysis and feature extraction of digitally modulated signal. In this paper, a kind of digital signal modulation identifier is designed based on wavelet. First, wavelet packet decomposition is operated on the digital signals, then according to frequency fragmentation characteristic of MFSK wavelet coefficients and amplitude variance feature of MASK wavelet coefficients, inter-class classification among MASK, MPSK and MFSK is implemented. Next, by analyzing the number of frequency fragmentations of MFSK wavelet coefficients and that of amplitude variances of MASK wavelet coefficients and that of zero-crossing layers of MPSK wavelet coefficients, intra-class classification among MASK, MPSK, MFSK is also realized. Simulation results suggest that the identifier has such a good anti-noise performance...