MAXIMUM-LIKELIHOOD-ESTIMATION OF CONSTELLATION VECTORS FOR BLIND SEPARATION OF COCHANNNEL BPSK SIGNALS AND ITS PERFORMANCE ANALYSIS

In this paper, we present a method for blind separation of co-channel BPSK signals arriving at an antenna array, This method consists of two parts: the maximum likelihood constellation estimation and assignment , We show that at high SNR, the maximum likelihood constellation estim ation is well approximated by the smallest distance clustering algorit hm, which we proposed earlier on heuristic grounds, We observe that bo th these, methods for estimating the constellation vectors perform ver y web at high SNR and nearly attain Cramer-Rao bounds, Using this fact and noting that the assignment algorithm causes negligble error at hi gh SNR, we derive upper hounds on the probability of bit error for the above method at high SNR. These upper bounds fall very rapidly with i ncreasing SNR, showing that our constellation estimation-assignment ap proach is very efficient, Simulation results are given to demonstrate the usefulness of the bounds.