Sensory Adds Speaker ID to Wake-up Words
Sensory, Inc., takes embedded speech processing another step forward by adding voice biometric-based speaker identification to its TrulyHandsfree(TM) voice activation technology. As a result, a wide variety of home and mobile electronic devices can be trained to “wake up” and respond to a specific person’s voice. For mobile professionals means higher levels of security when activating smartphones, laptops, desktops and tablets. There are equally important implications for owners and users of electronic devices in the home. As Sensory’s CEO Todd Mozer points out in this press release, the new software will make it possible to train a digital video player (or set-top box) to recognize the voice a specific viewers in order to recommend appropriate “favorites,” or block unsuitable programs.
As we’ve noted before, Sensory, Inc., has long been devoted to supporting TrulyHandsfree(TM) voice control of devices and media, especially in the automotive setting, where wake up words need to be distinguished from road noise or other random sounds in the background. Voice activation has been available on the Samsung Galaxy S2 for roughly a year now. In order to reduce the power consumption required to have the device constantly listening for its wake-up words, Sensory has made sure that the technology for the trigger words are “deeply embedded” in the devices hardware.
According to Mozer, the company “always had speaker verification on our chips.” However, to support the new offer, the company switched to a new HMM (hidden Markov model) based biometric engine in order to provide higher accuracy and lower false acceptance or false rejection rates. Sensory is demonstrating an alpha version of the system running on an Android device and using the passphrase “Hello Blue Genie”. Mozer trained the smartphone by repeating the phrase three times. Then he was able to unlock and activate the phone by saying the phrase, whereupon it said “Greetings Todd Mozer.” When i said the phrase after he had trained it, the phone woke up, but displayed the word “Rejected,” and I was unable to activate any of its features or services.
A beta version of the voice activation with speaker identification technology will be available under license from Sensory in the coming week. As the press release explains, details are available from firstname.lastname@example.org