摘要 |
A dynamic programming or DP matching system for speech recognition. Upon DP matching, cumulative distance is compared with a threshold value at every sampling time point of a speech pattern to thereby restrict the number of DP paths in succeeding matching processes. The number of DP paths remaining at each speech pattern sampling time point is monitored by a monitoring module for altering a threshold value so as to decrease the number of the DP paths. When DP path number becomes excessively large, the threshold is increased to thereby decrease the DP path number. Capacity of a DP data storing memory can be reduced while preventing the matching capability from being lowered.
|