News spread this week, thanks to a report from Time, that Amazon is working on a new feature that would give Alexa the ability to recognize an individual’s voice. In referring to sources “familiar with Amazon’s Alexa strategy,” the Time article mentions the voice signature feature has been under development since 2015. The challenge now is in strategically integrating it into Alexa devices like Amazon Echo.
The inability for Amazon Echo devices to distinguish between voices has been a noticeable feature for some time. At Opus Research’s Intelligent Assistants Conference last September in San Francisco, Dave Isbitski, Amazon’s chief evangelist for Alexa and Echo, acknowledged the topic of speaker identification but fell short of revealing any new developments about Alexa when it comes to voice identification.
Thanks to a surge in sales, Amazon and third-party developers have been rapidly expanding what the Echo line of devices can do in supporting more than 10,000 “skills.” The real potential for the voice ID feature would be the ability to distinguish among household members that have more than one Amazon account (or other skills providers) associated with a single Alexa.
Opus Research defines the Amazon Echo, as well as Google Assistant and Apple’s Siri and others, as “MetaBots” — embedded digital devices prevalent in our daily lives. A MetaBot operates on an individual’s behalf which puts them in the position of unhooking the direct link between brands and their customers.
It should be noted that rudimentary speaker identification (speaker ID ‘lite’) is already in use on iPhones with Siri, since 2015, as well as several other personal virtual assistants. With Amazon seemingly on the verge of unveiling voice identification features, these MetaBots are sure to become even more intelligent assisting us with secure, personalized services everyday.
Categories: Conversational Intelligence, Intelligent Assistants, Intelligent Authentication, Articles
GenAI Trust and Safety: Mitigating Risks and Enabling Solutions 
Opus Research Welcomes Ian Jacobs as VP and Lead Analyst 
Navigating a Future of Automation: Strategic Insights from NICE’s Analyst Summit 2024 
Verint Defines How to Stand Out in the GenAI Era
Amazon has pioneered the voice interface but whoever gains the ability to apply individual voice recognition would have a big advantage. I’m sure Google, Apple and Amazon are all working on it. Then there’s the case of teenagers whose voice change rapidly. Voice recognition that adapts over time?
Charles: You are absolutely right when you observe that “Google, Apple and Amazon are all working on it.” As for the changes in our voice characteristics as we age, you’ve identified another area that long-time voice biometrics companies have been addressing. Two things to think about: (1) much of the “science” of speaker recognition and speaker identification is based on setting thresholds regarding the level of confidence that is acceptable when judging whether a company or individual should be confident in the identity of another based on capturing spoken words. So there is some flexibility there (and voice is seldom, if ever, the only factor bring brought to bear to authenticate anyone. (2) It is quite common to continuously monitor “acceptable” voice prints and detect changes over time. As with any algorithm that governs decisionmaking (i.e. “accept/reject” a match between a voiceprint) it can be tweaked. In this case, you can use the algorithm to recognize and adjust for the impact of age. More likely, though, it is good advice for companies to periodically ask individuals to refresh their voiceprints, either explicitly, or in the background (with consent, of course).