Voice Biometrics at SpeechTEK 2011
Biometric Session Recaps...
The Monday biometric AM session on workflow and fraud prevention involved two customer use cases that utilized VoiceVault’s 5 random digit verifier presented by Nik Stanbridge and Julia Webb. In each case, having a biometric enabled the automation of processes which resulted in bottom line benefits. The first case was the management of remote home care attendants (landline based call-in system) which reduced ‘buddy punching’ fraud and increased automation for an estimated savings of 8%. The second case was a voice signature application which enabled Blue Cross Blue Shield of Kansas City to initiate policies over the phone versus previous paper based, postal mail applications decreasing unreturned applications significantly.
The Monday biometric PM session on more security for sensitive information involved a presentation on the current authentication space and where speaker biometrics fits in. This first presentation by iBiometrics’ Valene Skerpac (myself), highlighted the strengths of voice biometrics and where it is positioned within the inevitable shift to multi-factor authentication and wide reaching identity frameworks due to the weakness of passwords in the midst of a rapidly changing technology landscape. The second presentation by Percieve Solutions’, Roanne Levitt, highlighted the strengths of text independent voice biometrics compared to other biometrics as well as text-dependent voice biometrics methods. She discussed performance related techniques to handle channels and other factors such as aging or illness that may be relevant to certain user sets.
The Monday voice biometrics lab run by Judith Markowitz featured three vendors that allowed attendees to try their wares. Amongst other features demonstrated, Loquendo showed-off their identity and language classification software, VoiceVault showcased their verifier performing authentication with voice input via an iPad in a noisy environment and Recognition Systems showed their text independent verifier as well as their classifier software which recognized specific musical instruments and speaker gender.
Unfortunately, I missed the Monday sun-rise discussions on Voice Biometrics Basics but Tuesday I led the Best Practices discussion using an outline from the Best Practices document developed by the VoiceXML forum. We reviewed biometric technology and security fundamentals with a lot of discussion surrounding the various techniques employed in biometric engines. We touched upon standards/guidelines and recommended project life cycle approaches. Though the talk this year surrounded more around technology methods, interworking and performance there was also a review and some comments shared regarding user interface and acceptance issues.
Other Notable Moments...
A voice biometric login was included in a mock multi-modal demo created by David Attwater of EIG which demonstrated a phone call connecting to a smartphone browser GUI which prompted for the ID and voice password. In an instant, this demo opened up a new way of envisioning a mobile multi-modal speech enabled application.
Summary...
All in all, SpeechTEK covered voice biometrics well and its sessions were well attended with individuals from organizations close to implementation and beyond the ‘let’s think about it’ stage. In the wrap up, James Larson and Susan Hura acknowledged the popularity of today’s mobile applications and the fact that they rarely include speech as well as the lingering poor perception of outdated IVR speech applications still in use today. There was a call to re-invent, especially in the multi-modal space. As shown by the innovative mock demo described above, I think voice biometrics has great potential to help make this happen.