To my (limited) knowledge, when determining which survivor path to choose to produce an error-corrected output of a binary Viterbi decoder, the path with the smallest Hamming distance is the most likely estimation of the transmitted signal.

I have come across an example where I am presented with two survivor paths, both with the same smallest Hamming distance. Is there a method of determining which path to then pick? Or is there no method of doing so?

  • 1
    $\begingroup$ They are chosen randomly with an equal probability. $\endgroup$ – AlexTP May 9 '18 at 20:06
  • $\begingroup$ I am doing relatively short examples of this decoding method (10 received bits), with longer bit streams is it less common to have this similarity between survivor paths? $\endgroup$ – C. Faldon May 9 '18 at 20:10

Viterbi decoders are usually designed to take one of the two paths consistently versus the other in the case of a tie. This makes it a lot easier to do bit-exact non-regression testing of the implementation.

If the numeric precision of the input has enough resolution, the probability of a tie should be sufficiently small that it doesn't substantially impact performance to have this (presumably minuscule) bias. If there is not enough precision, then you will have non-trivial bias (and quantization noise), and such insufficient precision will result in performance degradation, regardless of whether the bias is randomized or consistent.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.