Training algorithm for speaker-independent voice recognition systems using HTK

Abstract

This paper presents the training algorithm by means of which everyone can develop their own speakerindependent voice recognition system. HTK toolkit is chosen as main tool for recognition process. Through this algorithm user may create voice recognition systems in a short period of time and automatically. Although a usershould prepare input data for training the recognition system. Test recognition systems, working on developed algorithm, have appeared to reach decent accuracy for the Belarusian language and show up themselves viable.