Summary

We have taken a look at how using a voice as an identifier is a behavioral biometric, in that it represents the manner in which a subject speaks. We then reviewed the many things that can influence the way we speak, such as medical conditions and aging. We examined the advances of VRT over the years and its similarity to speech recognition. We saw how to use statistical methods such as HMM, which helps recognize speech by looking for predictive patterns to change the state.

We reviewed how a voice is digitized and transmitted and learned how performance can be impacted by transmission errors and inferior equipment. We then stepped through the process of what happens from enrollment to matching. We compared text-dependent versus text-independent methods, and then compared matching methods that use either template matching or feature analysis. Finally, we saw how using VRT to quickly provide identification and authentication can streamline the overall experience and enhance...