Then we hold that revised estimate for talker A fixed and search for a better
estimate of talker B's speech.
Now we hold
our estimate of talker B's speech fixed and use the Viterbi algorithm to find the best estimate of talker A.
Not exact matches
Obtaining those
estimates still requires searching through a vast number
of possible sound combinations that the
talkers could have produced.
We form an initial
estimate of the likely speech
of each
talker.