Does anyone know if there's a good voice authentication library available? Or do I need to find a good voice recognition library and build from there?
There are several open source packages implementing speaker recognition. For example Mistral
http://mistral.univ-avignon.fr/
There are clones of mistral too
http://code.google.com/p/improved-mistral/
Speech Recognition is usually quite different from voice authentication (often called voice biometrics). Here is a good overview document - http://www.sans.org/reading_room/whitepapers/authentication/exploration-voice-biometrics_1436 says:
There are consultant companies such as J. Markowitz, Consultants who rank voice vendor software. These reports are often available for purchase online. One vendor’s implementation may be more suitable than another depending on the use case. Nuance (www.nuance.com) and Vocent (www.vocent.com) are two vendors who provide voice authentication solutions and who came up repeatedly in the research for the “Current Applications” section of this paper. Nuance has been rated the top voice authentication technology vendor by Celent Communications.12 Nuance claims a 96% or higher accuracy rate for speech recognition13, and Nuance’s Verifier product claims a “high” accuracy rate.
That is a good overview, but it is a little out of date. Look around for others such as:
- http://www.persay.com/technology-voice-biometrics.asp
- http://www.voxeo.com/biometrics/home.jsp
- http://www.nuance.com/landing-pages/enterprise/voice-biometrics/default.asp
- http://www.angel.com/products/voice-biometrics.php
- http://www.voicevault.com/voicevault-enterprise/
精彩评论