Single-channel blind separation of co-frequency modulated signals based on Gibbs sampler

Abstract

A Gibbs-sampler-based separation algorithm was proposed for single-channel co-frequency digital-modulated signals，which were received in non-cooperative ways.The probability density samples of unknown symbol sequences can be obtained by statistic method，and the computation complexity was not growing exponentially with the increase of channel order.With special emphasis on the separation algorithms of single pair of symbols，multiple pair of symbols and the tracking of channel responses.Moreover，the separation performance between Gibbs sampler and per-survivor processing(PSP) algorithm were analyzed and compared in detail.Simulation results show that，compared with PSP algorithm with L=4 for two co-frequency QPSK-modulated signals，Gibbs sampler can be nearly 17 times lower in computation complexity when the approximately performance is obtained.

Keywords

• single-channel blind separation;Gibbs sampler;PSP algorithm;paired carrier multiple access