A generalization of the fano metric and its effect on sequential decoding using a stack

In this work, we consider the application of sequential decoding to codes of practically short code length, such as a few hundreds bits, as the performance is conventionally deviated from the optimal maximum-likelihood (ML) performance. By re-formulating the original Fano metric as an equally weighted sum of two portions: the ML metric portion to previously received code bits, and an estimate portion on the ML metric of upcoming code bits, we found that the sequential decoding can be viewed as a special case of the Algorithm A. One can then vary weights on these two metric portions. Empirical investigation shows that by a proper choice of the weighting factor, employing the generalized (or weighted) Fano metric in sequential decoding can yield a markedly smaller bit-error-rate than the conventional (equally weighted) one, and the computational complexity is only moderately increased.