The 2017 Speech Industry Star Performers: Cobalt
Cobalt Comes Out of the Blue with Breakthrough Voice Control
Using speech applications in noisy environments, like warehouse floors, for example, is never easy. Separating the speaker’s voice from the chorus of other sounds challenges even the most sophisticated of systems.
But Cobalt Speech & Language just might have found the answer. Through a unique partnership launched earlier this year with VocalZoom, a supplier of optical sensors, the company is now poised to deliver a disruptive end-to-end solution for voice control in the connected car, on head-mounted devices used in industrial settings, and in access-control applications.
The Tyngsboro, Mass.–based start-up, which was founded in late 2014, is combining VocalZoom’s optical voice sensor technology with its own speech recognition engine to create a comprehensive voice-control development platform capable of delivering near-perfect performance even in noisy environments with other background speakers.
VocalZoom’s HMC optical sensors acquire and measure facial vibrations during speech and combine this data with the output from acoustic microphones to create an isolated, near-perfect reference signal regardless of noise levels.
Preliminary tests of Cobalt’s speech recognition engine with the VocalZoom sensor under noisy conditions demonstrated an almost 60 percent improvement in performance.
“I’ve been working to advance the state of the art in speech recognition technology for more than two decades and have achieved more performance improvements working with VocalZoom’s HMC sensor in two months than would typically be seen in two years with conventional approaches,” said Cobalt’s founder and CEO, Jeff Adams—who previously built and managed speech and language software groups at Nuance Communications, Yap, and Amazon—in a statement. “I believe we are now poised for unprecedented breakthroughs and a new realm of voice control product possibilities using this combined solution. It will open the next chapter in how we control our increasingly digital world.”
The Cobalt-VocalZoom partnership is also expected to yield a product development platform that could potentially speed the time to market needed for similar products.
Rammy Bahalul, vice president of sales and business development at VocalZoom, notes that Cobalt has “been at the forefront of so many of today’s speech recognition technology innovations.” The new platform, he says, “will give consumer electronics manufacturers worldwide a platform for significantly improving their current products, with no changes required to their existing speech architecture or noise reduction algorithm.”
Cobalt’s advances in speech recognition are also changing the way high school and college students learn foreign languages. Vista Higher Learning, a Boston-based publisher of world language learning materials, in late 2016 launched the Verification Automatic Speech Recognition (VASR) engine that incorporates Cobalt’s technology. VASR is used to determine how well students’ speech compares to specific, predetermined words or phrases.
The first implementation of VASR was applied to Vista’s Portales product, an introductory Spanish program in use at more than 100 colleges and universities. Portales prompts learners to state a word or phrase; it then records the speech and provides real-time feedback with an emphasis on vocabulary recognition and speaking skills. In all, the program features more than 300 activities covering approximately 4,000 unique Spanish words and phrases.