Abstract
We design low-complexity bi-directional (forward and backward) modified Soft-Output Viterbi Algorithm (SOVA) decoders to exploit source memory from binary Markov and hidden Markov model (HMM) sources, respectively, for decoding in serially concatenated continuous phase modulation (SCCPM) systems. The SCCPM system consists of a recursive systematic convolutional (RSC) code serially concatenated with an M-ary continuous phase modulation (CPM) signal, forming an iterative decoding structure where SOVA serves as the soft-input soft-output (SISO) decoder for both the CPM and RSC code components. The modified RSC SOVA decoders integrate soft information derived from source statistics (Markov or HMM characteristics) with the extrinsic information from the CPM SOVA, and the integrated soft information is used as refined a priori information for the RSC SOVA. The updated extrinsic information from the modified RSC SOVA is then exchanged with the CPM SOVA in subsequent iterations. Simulation results demonstrate that the proposed methods can effectively exploit the source statistics, thereby significantly enhancing error performance and accelerating decoding convergence in iterative decoding.

This publication has 35 references indexed in Scilit: