作者: M.J. Roach , J.D. Brand , J.S.D. Mason
关键词:
摘要: This paper gives an insight into biometrics used for speaker recognition. Three different are presented, based on: acoustic, geometric lip, and holistic facial features. Experiments carried out using a corpus of the DAVID audio-visual database. Recognition accuracy is found to be similar in 2 domains. The visual feature on method signature coding contour lips mean dynamic signature, capturing motions face during spoken utterance. Physical (static measurements) demand only small model sizes, perhaps just single template, therefore require less training data. Conversely behavioral contain more variation