ICA-Based Lip Feature Representation for Speaker Authentication

Abstract

Compared with some "static" biometrics such as human face and fingerprint, person authentication based on lip movement has the advantage of incorporating "dynamic" features which contain rich information indicating the speaker identity. This paper proposes a new lip feature representation and analyzes its discrimination power for person authentication. Since the original lip features are usually of high-dimension, the independent component analysis (ICA) is adopted for dimension- reduction and discriminative feature extraction. Hidden Markov model (HMM) is then employed as the classifier for its superiority in dealing with time-series data. Experiments are carried out on a database containing 40 speakers in our lab. By analyzing the experimental results, detailed evaluation for various lip feature representation is made and 98.07% accuracy rate in speaker recognition and 2.31% EER in speaker authentication is achieved using our lip feature representation.