Efficient statistical parallel interference cancellation for DS-CDMA in Rayleigh fading channels

We present a reduced-complexity parallel interference cancellation (PIC) technique, in which PIC is performed only on "unreliable" bits (blocks). We term our technique statistical PIC (STPIC), since the decision to employ PIC for any user signal is based upon received signal statistics. Here we propose two different methods: one is bitwise (instantaneous) STPIC (ISTPIC), in which IC is performed only on the bits whose absolute log likelihood ratio (LLR) is below a pre-selected threshold; the other method is block STPIC (BSTPIC), in which IC is performed only on the blocks whose average signal-to-noise-plus-interference ratio (SNIR) is below a pre-selected threshold. Both LLR and SNIR statistics reduce to quantities simple to obtain. We show that STPIC can achieve performance equivalent to conventional full PIC (FPIC) and other partial PIC techniques in flat Rayleigh fading channels, but with reduced computational complexity. We also show our proposed ISTPIC can achieve performance better than FPIC in good channel conditions (e.g., AWGN), still with reduced computational complexity. The ISTPIC performance gain over conventional FPIC is larger in conditions of larger loading factor where efficient IC techniques are more likely to be needed