Understanding speech in the midst of background noise has long reigned as one of the most difficult obstacles to overcome in hearing technology. However, a computer algorithm developed by team of engineers at Ohio State University, led by DeLiang Wong, professor of computer science and engineering, could be the linchpin to break through that cacophony for the hearing-impaired.
Wang says the algorithm is unique, because it uses a “machine learning” technique. Along with doctoral student Yuxuan Wang, they’re training the algorithm to separate speech by exposing it to different words in the midst of background noise. A “deep neural network” does the processing. The network “learns” through a deep layered structure inspired by the human brain.
Initial tests performed by Eric Healy, director of Ohio State’s Speech Psychoacoustics Laboratory and doctoral student Sarah Yoho involved 12 hearing-impaired volunteers. They removed the volunteers’ hearing aids and then played recordings of speech obscured by background noise—stationary noise (a constant noise, e.g. the hum of an air conditioner) as well as noise from voices, or babble, in the background—over headphones. They asked the participants to repeat words they heard, and then re-performed the same test with recordings, but processed with the algorithm to remove background noise.
Results showed that hearing comprehension with algorithmically altered background babble improved 25% to 85% on average. The algorithm improved stationary-noise comprehension from an average of 35% to 85%.
Though the initial tests focused on pre-recorded sounds, the researchers plan to refine the algorithm to better process speech in real time. It’s hoped that the technology will usher in the next generation of digital hearing aids that possibly reside inside smartphones. The phones would perform the computer processing and then wirelessly broadcast the enhanced signal to miniaturized earpieces.
Several patents are pending on the technology. Researchers are working jointly with hearing-aid manufacturer Starkey, among others, to further develop the technology. It’s currently being commercialized and is available for license from Ohio State’s Technology Commercialization and Knowledge Transfer Office.