Research On Techniques In Automatic Modulation Classification For Digital Communication Signals

Modulation scheme is one of the most important characteristics used to distinguish communication signals. So called automatic modulation classification (AMC) specifies that, given a received communication signal with unknown modulation information, the modulation type and relevant modulation parameters of the communication signal can be identified without any manipulation. The research of automatic modulation classification is of significance in both militarily and civilian applications.Research on automatic modulation classification has been carried out for at least two decades, and two general classes of AMC algorithms can be crystallized, likelihood-based (LB) and feature-based (FB). The former usually suffer from computational complexity, and are sensitive to model mismatches. This limits their practical application. The FB approach, on the other hand, is simple to implement, with near-optimal performance, when designed properly, and robust to model mismatches. In general, AMC is an active and challenging task so far.The focus of our study in this dissertation is on the automatic modulation classification of digital communication signals. The classification algorithm with robust performance, low computation complexity and high practical is expected achieved. In the dissertation, a comprehensive survey of current AMC techniques in a systematic way is provided. For different algorithms, the main characteristics are described and the bottle-necks are highlighted. Based on it, some further research topics are point out. The main contribution of this dissertation can be summarized as follows:Spectrum line features of digital communication signals are analyzed. With the nonlinearities applied to the complex envelope of the digital communication signals, signals of different modulation schemes always manifest themselves in spectrum line feature. A complete theoretical analysis for the square and quartic spectrum line feature is carried out. The spectrum line's existence, position and amplitude are deduced. Computer simulation results indicate that the spectrum line can properly be extracted even in the serious noisy and multi-path fading environments.A novel modulation classification approach for constant modulus digital modulation signals is proposed. The proposed approach doesn't need the prior knowledge of symbol rate, carrier phase, timing recovery, and so on, and can classify the linear and nonlinear digital modulation schemes. Simulation result illustrate that the proposed approach can achieve satisfying classification performance even when the SNR is lower, which shows the proposed approach is feasible and practical.A modulation order recognition algorithm for M-ary CPM signals is proposed. The relationship of spectrum line feature between M-ary and binary CPM signals is deduced. Based on it, the modulation order can be recognized correctly for M-ary CPM signals.We investigated the application of higher order cumulant in AMC. Estimation error performance and affected factors of higher order cumulant are reviewed, and the strategy of selection of cumulant order is pointed out.A new approach to classification of higher order PSK/QAM signals in multipath fading environments is presented. The proposed approach, in which the two-step equalization strategy and higher-order cumulants based classifier are adopted, can effectively classify the PSK and higher-order QAM signals. The performance of proposed approach is evaluated by the computer simulations, which shows it has better classification ability.As an auxiliary technique, a joint algorithm for blind equalization and modulation index estimation of full-response CPM signals is proposed. The new algorithm, based on the analysis of solutions to the traditional CM criterion with input of CPM signals, can simultaneously achieve blind equalization of channel and estimation of unknown modulation index. Simulation results illustrate the performance of the algorithm.We proposed an effective modulation classification algorithm aimed at single-channel mixed communication signals. The diversification of cumulant and spectrum line feature of mixed signals is analyzed. Then they are utilized to active the classification of common digital modulation types. This modulation classification algorithm can help supply model information for PCMA system models.