On Detection Of Mppsk Signals Based On SVM Multi-Classifications

With the development of information society, the demand for wireless multimedia communications continues to increase and the modern wireless mobile communications need broader bandwidth, the limited spectrum resources becomes increasingly scarce. Ultra narrow band modulation is a technology with high spectrum efficiency, which can solve the problem of spectrum shortage. This paper researches the usage of support vector machine (SVM) in MPPSK signal detection based on achievements of the EBPSK(Extended Binary Phase Shift Keying, EBPSK)/MPPSK(M-ary Position Phase Shift Keying, MPPSK) modems and the usage of SVM in the EBPSK signal detection.Firstly, the modulation theories of EBPSK and MPPSK are introduced, and the theory of the key demodulation technology, namely impacting filter, is elaborated using MPPSK as an example. Then after analyzing the advantages and disadvantages of traditional signal detection methods and the difficulties of MPPSK signal detection, SVM is introduced to the high efficiency communication systems, and its basic principle is illustrated. To compare the signal detection efficiency among traditional methods and SVM algorithm, the BER of EBPSK system is simulated.Secondly, on the theoretical basis of existing SVM multiclass algorithms, similar dichotomy SVM algorithm is proposed which can guarantee the detection efficiency of low order MPPSK modulations and deduce the algorithm complexity at the same time. Then the existing MPPSK signal detection methods are generalized with their performance compared to similar dichotomy SVM via simulation experiments. On this basis, LDPC (Low-density parity-check code) CODEC is accomplished and decent coding gain is achieved using special CODEC mechanism.Then, parameters which influence the performance of SVM algorithm are selected through experiments by use of the LIBSVM package, and band-pass filter of the receiver is also designed in detail. The simulation results illustrate that the LIBSVM has superior detection efficiency and low complexity, and the impacting signal is more suitable for SVM network training and detection efficiency is better after the receiving band-pass filtering.Lastly, the usage of SVM in high order MPPSK modulations is discussed with similar dichotomy SVM as the detection method and modulation array algorithm as the feature selecting method, and the algorithm scope is given then.