Single-channel blind separation of Two QPSK signals using per-survivor processing

In this paper, we consider a blind separation problem for single-channel mixtures of two QPSK signals, which may be slightly different in carrier frequencies. We present the joint estimation of channel impulse responses (CIR) and data symbols for both sources. The proposed approach uses a per-survivor processing (PSP) algorithm which is efficiently implemented by combining the simple least mean square (LMS) algorithm and decision feedback sequence estimation (DFSE). We initialize the CIR resorting to amplitude and phase estimation to ensure and accelerate convergence. We also discuss the problem of timing. Simulation results are provided to illustrate the effectiveness of the proposed algorithm.