Audience, a California-based startup, has made a noise-canceling chip for cell phones that could also improve voice-recognition systems.
In 1989, Lloyd Watts headed to Caltech to get his PhD in electrical engineering and join microelectronics pioneer Carver Mead’s effort to understand the human brain. Watts’s task: studying the mechanics of the human auditory pathway.
Over the next two decades, Watts’s work morphed into a startup called Audience, which amassed $45 million in venture funding. This year, it rolled out its first product: a chip for mobile phones that cancels out a wide range of background noises–even loud public-address systems in airports. The technology could also improve the performance of voice-recognition systems. Today’s customer-service and voice-command systems are increasingly sophisticated, but they can still be defeated by background noise.
Working first at Caltech and later at Microsoft cofounder Paul Allen’s now-defunct Interval Research, Watts and other researchers built computer models of the inner ear, or cochlea, including the fine hair cells that detect different frequencies of sound (some detect high frequencies, others low).
At Audience, Watts and his team were able to distill these models into algorithms that could be wired into a chip. The system models the frequency, duration, and volume of sounds coming into a cell-phone microphone. Then it uses that information to group sounds together according to source. Finally, it deletes the audio streams that it has identified as noise, leaving only the sound of a speaker’s voice.
The technology “literally replicates the human hearing system,” Watts says. “We directly studied the biology, working with auditory neuroscientists and doing simulations of the cochlea.” The company says that its system can suppress 25 decibels of noise, versus 6 to 12 decibels with conventional methods. Audience is now selling voice-processor sample chips, for $5 to $7, to makers of cell phones. Its first customer, Sharp, is using them in an NTT DoCoMo phone in Japan.
Stephen Ohr, an analyst at market research firm Gartner, says that companies including AMI Semiconductor, Qualcomm, and NXP are all developing similar technologies. While Audience has a head start, he says, the market battle will ultimately be decided by questions such as price and manufacturing quality.
Location: Mountain View, CA
Product: Voice-processor chip that sharpens voice quality and filters out noise in cell-phone calls
Founder: Lloyd Watts
CEO: Peter Santos
Number of employees: 50
Funding amount: $45 million
AI is here.
Own what happens next at EmTech Digital 2019.
The Best of the Physics arXiv (week ending December 15, 2018)
This week’s most thought-provoking papers from the Physics arXiv.
At least 15 central banks are serious about getting into digital currency
Digital cash may soon start replacing the physical kind.
New York’s new wage law for Uber drivers is a lesson for cities around the world
Alex Rosenblat rode along with Uber drivers for 5,000 miles. We talked with her about the huge impact the law could have.