next up previous
次へ: Forward decoding ( New 上へ: New word spotting algorithm 戻る: Ergodic HMM connecting Word

Decoding for optimal state sequence

Given the Ergodic HMM parameters and speech data, we consider the problem of estimating the optimal state sequence. To resolve this problem, the Viterbi decoding is a formal technique for finding the single best state sequence. Word sequences are directly estimated to obtain this optimal state sequence.

Now, we propose the new decoding algorithm. We call this algorithm the forward decoding. It is described in the following.



Subsections

Jin'ichi Murakami 平成13年1月19日