Biometric person authentication (1)
A number of biometrics have been proposed including, voice, face, fingerprints, iris, infrared facial and hand vein thermograms, ear, gait, keystroke dynamics, DNA, signature and acoustic emissions, odor, retinal scan , hand and finger geometry.
A voice signal is typically degraded by the microphone, communications channel, and digitizer characteristics.
Matching strategies: HMMs, vector quantization or dynamic time warping.
Types of speaker authentication: text-dependent, text-independent, and language-independent.