Demodulation of a M-Ray Position Phase Shift Keying System Using Multi-Class Support Vector Machine Classification

Abstract

In this paper, we introduce a new approach for nonlinear demodulation based on multi-class support vector machine (SVM) classification. We propose to measure the performance of this demodulator with different M which is the parameter of M-ray position phase shift keying (MPPSK) modulation, and compare with other demodulation technique. During demodulation, a few sampling points are chosen for multi-class SVM training and testing, which can reduce the complexity of system. Simulation results show that this new approach significantly outperforms the method of using Phase Locked Loop (PLL) demodulation by 10dB, and also better than Back Propagation Artificial Neural Networks (ANN-BP) classification demodulation. With the growth of M, the data rate increased and the performance become a little worse, but less bit SNR is used to achieve the same Symbol Error Rate (SER) as small M. So, it is an effective method to get better performance by using multi-class SVM classification technique for demodulation in MPPSK system.