APPARATUS AND METHOD FOR SPEAKER RECOGNITION

摘要

PURPOSE: A speaker verification device and a method thereof producing likelihood score of a specific combination with a minimum complexity are provided to calculate likelihood value about a characteristic combination comprised of a plurality of characteristic by minimum complexity and calculation. CONSTITUTION: A characteristic extracting unit(200) converts a received voice signal into a frequency area and extracts a N numbers of characteristic from the converted voice signal. A matching unit(300) calculates a likelihood score of extracted N characteristic and calculates a likelihood score of M characteristic and recognizes a speaker based on a likelihood score of a characteristic combination.