摘要 |
PROBLEM TO BE SOLVED: To improve recognition rate by increasing the accuracy of matching. SOLUTION: A phoneme string sent from a phoneme recognition part of a speech recognition device is supplied to a phoneme string input process 1. Next, the part where a same phoneme succeeds is extracted from the phoneme string inputted to the phoneme string input process 1 by a same phoneme continuance extracting process 2. A phoneme judgment process 3 judges whether or not the same phoneme extracted by this same phoneme continuance extracting process 2 match with the phoneme in dictionary templates being compared now. The phoneme judged as matched in this judgment process 3 is conditionally brought out of a tilt limit, and inputted to a pass transition permission process 4 for tilt '0' and is permitted to pass with the tilt '0' for being matching-processed.
|