Modulation Classification Of Digital Signals Based On Wavelet-ridge

Modulation type is one of the most important characteristics used to distinguish different communication signals. Given a received signal, the objective of modulation recognition is to identify the modulation type and estimate the modulation parameters without any priori knowledge. This dissertation is focusing on a study of instantaneous characteristics extraction based on wavelet-ridge and its application in Inter-class and Intra-class classification of digitally MPSK/MFSK modulated signals. Based on the achievements of the predecessors in this field several innovative algorithms are proposed and implemented. The main work of this thesis can be summarized as follows:Based on analysis of wavelet-ridge theory a point of view is proposed that in wavelet-transform-domain (WTD) there are not distinct wavelet-ridges always existing for all kinds of digitally modulated signals, thus preprocessing of the signals before extracting the wavelet-ridge is necessary. A criterion is given which is used for making a judgement of whether preprocessing is necessary. If the requirements of the criterion is not met, interpolation and upward-frequency-moving is needed before successive wavelet transform. The effects of parameters selection of Morlet wavelet on the resolution and precision of wavelet-ridge is also investigated.To solve the disadvantages of the wavelet-ridge extraction algorithms based on modulus maximum of wavelet transform and phase information, a modified is proposed, which can speed up the algorithm and improve the anti-noise performances of wavelet-ridge extraction. A comparison with the conventional wavelet-ridge extraction algorithms based on phase information and modulus maximum of wavelet transform is implemented, which proved the feasibility and practicability of the new algorithm, especially under low SNR.The Inter-class and Intra-class classifications of digitally MPSK/MFSK modulated signals with pulse shaping are investigated based on instantaneous characteristics extracted from wavelet-ridge. Estimation approaches of symbol rate and bandwidth of the signals based on wavelet-ridge are proposed. And a modified automatic identification algorithm based on symbol hopping for low SNR situation is proposed.Finally, the overall scheme and modular structure of an automatic identification software system is provided and discussed, and its implementation is described in detail.Simulations results have proved the validity and feasibility of the methods proposed in this thesis.