Exclusive: Amazon Developing Advanced Voice-Recognition for Alexa
Amazon is working to make its Alexa a better listener.
The Seattle-based technology giant has been developing a feature that would allow the voice assistant that powers its Echo line of speakers to distinguish between individual users based on their voices, according to people familiar with Amazon's Alexa strategy. The sources declined to be identified by name because they are not authorized to talk about the company’s future product plans. An Amazon spokesperson declined to comment.
Alexa, like Apple’s Siri or Microsoft’s Cortana, can interpret and respond to voice commands such as “How’s the weather?” or “What movies are playing tonight?” So far, though, none of the mainstream voice-enabled smart speakers have been able to distinguish who in a household is asking for something. Amazon’s new feature would match the person speaking to a voice sample, or “voice print,” to verify a person's identity, according to a source. A primary account holder would be able to require a specific voice print to access certain commands. A user would, for example, be able to set it so that a parent's voice would be required to make a credit card purchase or turn on the coffee machine through the Echo.
Read more: The ultimate guide to the Amazon Echo
Amazon has been developing this capability, internally called Voice ID, since at least the summer of 2015. It remained on the Alexa roadmap as recently as late last summer, but it's unclear when or if the feature will launch. The underlying technology has been completed; it's just a matter of integrating the feature into Echo products, one of the sources claimed. A number of factors could delay or the feature, including privacy concerns, which have become a greater issue for the company. Amazon recently refused to turn over voice records from an Echo user to Arkansas police investigating a murder which took place last November, saying that such recordings should be protected under the First Amendment.
The technology could provide another boost to Amazon’s surprise-hit Echo line of devices. Since its launch in 2014, Amazon and third-party developers have been rapidly expanding what the smart speaker can do. Alexa now supports 10,000 “skills”—such as compatibility with services from Uber and Capital One— up from 7,000 in January and 3,000 in September. Amazon doesn't share sales figures for its Echo devices, but Morgan Stanley estimates that 11 million units had been sold as of Dec. 2016. Alexa also appeared in devices ranging from home robots to car infotainment systems at this year's CES conference.
Amazon already allows Echo owners to set up multiple profiles and jump between them, but the user must say "switch accounts" or use the Alexa app to do so. There's also a setting in the Alexa app that requires a four-digit authentication code to be said out loud in order to confirm credit card purchases. But no feature for identifying authenticated users based on a voice print is currently available.
Adding Voice ID could make the experience of sharing an Echo with multiple family members or roommates more seamless. It may also prevent scenarios like the one Megan Neitzel found herself in this past January, when she was unaware that her six-year-old daughter Brooke ordered an expensive doll house and a four-pound cookie tin until it arrived at her home. Beyond shopping and authentication, giving Alexa knowledge about who is speaking may allow it to play music that’s more closely aligned to a specific user’s tastes. ..Continued.. by Lisa Eadicicco