Turbo codes optimization via trace-bit injection and selective puncturing

This paper looks at the problem of optimizing a given parallel concatenated convolutional code (PCCC) via an iterative process of trace-bit injection which minimally reduces the code rate, followed by selective puncturing that allows recovery of the rate loss incurred during the trace-bit injection. The technique allows for optimization of any linear PCCC. The optimization is in the sense of trimming the distance spectrum of the overall PCCC via elimination of the low distance and/or high multiplicity codewords from the output space of the code, using a cost function closely related to the asymptotic bit error probability of the code. This improves the performance of the code everywhere, but the main impact of the optimization process is a reduction in the error floor of the PCCC.