Metamaterials assist in ‘cocktail party’ voice recognition

Engineers at Duke University have developed a new type of sensor that enhances a computer’s ability to identify individual sounds that are overlapping, helping address the so-called “cocktail party” problem of voice recognition software. 

The device, described in the Proceedings of the National Academy of Sciences, is shaped like a large honeycombed pie with dozens of interlocking slices. The depth of the openings varies in each slice of honeycomb, meaning sound passing over each section has a unique pattern.

“The cavities behave like soda bottles when you blow across their tops,” said Steve Cummer, professor of electrical and computer engineering at Duke. “The…depth of the cavities…affects the pitch of the sound they make, and this changes the incoming sound in a subtle but detectable way.”

When sound passes over the acoustic metamaterials, it gets slightly distorted depending on the depth of the cavities. The sound is picked up by a microphone on the other side of the device, then transmitted to a computer that is able to separate the overlapping noises based on the unique distortion patterns.

According to the team at Duke, the six-inch wide prototype used in testing had a 96.7 per cent success rate in identifying sounds from three different directions. If the technology can be successfully miniaturised, the researchers believe it has a wide range of potential applications, from consumer electronics and medical sensing devices, to hearing aids and cochlear implants.

“We’ve invented a sensing system that can efficiently, reliably and inexpensively solve an interesting problem that modern technology has to deal with on a daily basis,” said Abel Xie, a PhD student in electrical and computer engineering at Duke and lead author of the paper. “We think this could improve the performance of voice-activated devices like smart phones and game consoles while also reducing the complexity of the system.”    

“With the extra information, it should also be possible to improve the sound fidelity and increase functionalities for applications like hearing aids and cochlear implants. One obvious challenge is to make the system physically small. It is challenging, but not impossible, and we are working toward that goal.”