On Recent Advancement of Automatic Speaker Authentication

Speaker authentication, which concerns with authenticating a person's identity via voice, is addressed in this paper. Two approaches to speaker authentication, speaker verification and verbal information verification (VIV), can be formulated. The traditional approach of speaker verification attempts to verify a speaker's identity based on his/her voice characteristics while the newly proposed VIV approach verifies a speaker's identity through the verification of the content of his/her verbal information. In this paper, we first introduce the technology components, then present the recent advancement in a VIV system and a fixed-phrase speaker verification system with a stochastic matching algorithm developed for channel compensation. Based on our lab evaluation, both systems are ready for real-world applications.