Signal detection with belief propagation in Faster-than-Nyquist signaling

Faster-than-Nyquist (FTN) signaling is one of the approaches that improve spectrum efficiency by transmitting symbols faster than the Nyquist rate. However, it suffers from inter-carrier interference (ICI) and inter-block interference (IBI). Thus, the ICI and IBI canceller is needed at a receiver. In this paper, signal detection based on belief propagation (BP) is applied to FTN symbol detection. The BP algorithm provides very high performance in signal detection with large scale random factor graphs. However, when we apply the BP algorithm to the FTN signal detection, the factor graph has many short loops locally, and randomness cannot be found in the edge strength distribution. Thus, proper convergence is not always expected. To improve the convergence performance in such an environment, we propose a technique to scale a posteriori LLR values before channel decoding. The simulation results show that the error floor is significantly reduced by this LLR scaling.