Performance of an MLSE-based early stopping technique for turbo codes

A crucial issue in the practical application of turbo codes is decoding complexity. A common approach to reducing decoding complexity is to stop the iterative decoding process early. It is demonstrated that an early stopping technique based on maximum-likelihood sequence estimation (MLSE) can significantly reduce average decoder processing. Specifically, it is shown that an 8-state turbo code with 512 data bits and a nominal code rate of 1/3 can be decoded in less than 2 iterations on average when E_b/N₀ ≥ 2 dB (packet error rate ≈ 10^-7). This reduction is achieved with virtually no loss in error rate performance relative to performing a large fixed number of iterations.