Modulation Recognition Methods Of Multiple Co-channel Signals

Modulation recognition of communication signals is an important research topic in both communications and non-cooperation communications. In the last decade, with the rapid development and wide application of radio communication technology, the signal environment is getting more and more complex, as a result, more and more challenging problems are encountered by the modulation recognition community. Among of them, the modulation recognition of multiple co-channel signals is one of the most difficult problems. Unlike modulation recognition of single signal, the modulation recognition of multiple signals is to identificate the modulation type of each signal from the mixed waveform of multiple unknown signals corrupted by additive noise. Modulation recognition methods presented for single signal case are not valid for multiple signals' case, including decision theory based and statistic pattern recognition based methods.The objective of this dissertation is focused on the modulation recognition methods for multiple co-channel signals. According to the signals' different distribution in frequency domain and spatial domain, two fundamental approaches for modulation recognition of multiple co-channel signals are proposed, namely, signal separation based and direct feature extraction based methods respectively. The main content and contributions of this dissertation are summarized as follows:(1) The problem of of modulation recognition for multiple co-channel signals with two data acquisition models, single wide band channel and multiple narrow chnannels, is discussed, and then two fundamental modulation recognition approaches for multiple co-channel signals are presented.(2) Based on the analysis of the feature extraction, a general framework of feature analysis is proposed for modulation recognition, then the featue extraction, the reduction of feature dimension, the construction of multi-feature vector and multi-feature set are discussed, respectively.(3) Based on the detailed review of classifier design, a kind of multi-feature and multi-label based classifiers is proposed to accommodate the fact of large SNR dynamic range and the requirement of multiple feature extraction in modulation recognition tasks, and then combined classifiers are also introduced.(4) Signal separation based modulation recognition of multiple co-channel signals is detailed investigated, a novel semi-sphere antenna array as a spacial filter is proposed for signal separation, and then a new AR model parameters based modulation recognition method for separated signals is presented.(5) Finally, by employing the feature of each signal directly extracted from the GAR parameters of the observed co-channel signals' data, a novel modulation recognition method for multiple co-channel signals is proposed, and numerous computer simulations are also performed for the two and three signals' case with different overlap percent in frequency domain. The simulation results show that this method is valid for modulation recognition of multiple co-channel signals.