Isolated Word Recognition for Thai Language using Fuzzy K-Means

摘要

This paper describes the implementation of a Thai digit recognition system. The digit classification scheme is based on fuzzy k-means algorithms to partition the training data for each class into several clusters, and the cluster center with the radius is computed. Then the unknown Thai digit is belongs to any corresponding cluster, we simply take the nearest cluster to represent it. The experimental results show that fuzzy k-means algorithms have a good performance to classify the data sets. The tests were done on an isolated word speech database of Thai with 20 and 12 utterances. The accuracy rate of the speaker-dependent case is 94 %, and the speaker-independent is 90 % respectively.