Variable sized- Clustering apparatus and method for better discrimination of the inter-speaker variation Speaker Verification apparatus and method based on new VQ/HMM technique

摘要

PURPOSE: An apparatus for generating a code book for improving discrimination between speakers and a method therefor, and an apparatus for recognizing speakers in a code book combination method using the same and a method therefor are provided to improve a recognition rate of verifying the speakers by emphasizing individuality of the speakers. CONSTITUTION: A center value setting element sets up one center value to all sampled voice vectors(S601). A detecting element detects a cluster having the greatest distribution out of all clusters(S602). A splitting element moves the center value of the cluster having the greatest distribution by considering the distribution to split the center value into two(S603). A collectivizing element measures euclid distances of split clusters with all input vectors for collectivizing a member vector of the cluster having the shortest distance(S604). A center point updating element updates a center point of the cluster through the member vector of each cluster for generating a personal code book(S605-S607).