August 2, 2013 -
Sensory focuses on embedded solutions and has been working with speech recognition and speaker identification for over 20 years.
I had a chance to chat with Bernie Brafman, VP of Business Development for Sensory, to talk about voice biometrics, the company’s current market focus, as well as what to expect from Sensory down the road.
Brafman is a veteran of the speech technology industry. He’s been working for Sensory as VP Business Development for the last year and a half, though this isn’t his first time at the company. He was an early hire in 1999, but after some restructuring, left the company in 2001 and continued to work with speech technologies. He holds a degree in engineering, but made the transition to sales and marketing early in his career.
“Where we are at today is very focused on mobile and focused on embedded software, either at the OS level or running on a chip,” Brafman said, noting the company’s recent software integration in the new Samsung Galaxy S4 smartphone, which enables users to activate the phone using voice triggers.
According to Brafman, Sensory’s edge in the marketplace is the fact that its systems are embedded, requiring no cloud or network connectivity – improving security and functionality for the user, he says.
“We are continually improving and expanding our performance and capabilities. One very important development is the availability of our TrulyHandsfree Voice Control on DSP IP cores and ICs from companies such as Tensilica, CEVA, Conexant and Wolfson that allow for very low power consumption in an ‘always on, always listening’ mode.”
As a long-time player in the market, Sensory has a large focus on consumer products and has some of the most-used technology today in terms of speech and speaker recognition.
“Our focus has really been on consumer since the start,” Brafman said. “We actually make a low-cost microcontroller that can run consumer products as well as run our speech technologies. We’re on the fifth generation of those devices. They are in toys, clocks and all kinds of consumer electronics that have voices in them.”
BiometricUpdate.com recently took an extended look at voice biometrics as a modality, including some early examples, as well as an exploration of potential developments and innovations. Read the full feature here.
Though the company’s focus isn’t necessarily on identification, Brafman says its capacities are significant.
“Our equal error rate has been measured at one percent in quiet conditions, and around 7 or 8 percent in noise. I think we have best in class in terms of that, particularly if you factor in a passphrase only you know — then it goes to a tenth of a percent.”
Having been working in speech technologies and with Sensory throughout his career, Brafman says he’s fortunate to have watched the evolution of speech and speaker recognition technology so closely.
“If I thought the stuff we were doing in 1999 was cool, the stuff we’re doing today is unbelievable.”