REAL TIME SPEAKER RECOGNITION SYSTEM AND METHOD USING VOICE SEPARATION

摘要

PURPOSE: Real-time speaker recognition system and method using voice separation are provided to confirm identification of a client without stopping of consultation work by searching and comparing a separated client's voice with a pre-established voiceprint database in real time while talking to the client on the phone. CONSTITUTION: A real-time speaker recognition system connects client and consultant on the phone, separates and records a voice signal of the client in real time (S110, S120). The system generates voiceprint information of the client using the recorded voice signal (S130). If the voiceprint information of the client is not registered in a database of a speaker recognition server, the system registers the generated voiceprint information to the database (S140, S150). If the voiceprint information is registered in the database, the system compares the generated voiceprint information with the existing voiceprint information registered in the database (S160). [Reference numerals] (AA) Start; (BB) No; (CC) Yes; (DD) End; (S110) Step of connecting a client and a consultant on the phone; (S120) Step of recording a voice signal of the client in real time; (S130) Step of generating voiceprint information of the client using the recorded voice signal; (S140) Voiceprint information exists?; (S150) Step of registering the generated voiceprint information to the database; (S160) Step of performing to certify by comparing the generated voiceprint information and the existing voiceprint information