Cocktail-party help for the hearing-impaired


OSU researchers create computer program that filters out background noise


By Dean Narciso
Clinking glasses; giggling children; and layers of music, laughter and speech weave a blanket of babble for the hearing impaired, making conversation futile and holiday parties a chore.
It’s called the “cocktail party problem” — the inability to filter unwanted background noise to focus in on one voice. Solving this problem is considered the Holy Grail of hearing technology.
“Satisfaction with hearing aids is notoriously low, and (background noise) is probably the primary reason,” said Eric Healy, the director of the Speech Psychoacoustics Laboratory at Ohio State University.
But Healy and DeLiang “Leon” Wang, a professor of computer science and engineering, say they have developed a computer program that might help solve this problem and might one day lead to better digital hearing aids.
The program uses mathematical algorithms and computerized neural networks to remove unwanted noise, kind of in the same way you might remove component colors from a can of brown paint.
“It’s very easy to add red plus green plus blue, and you know what the result will be,” Healy said. “It’s very hard to say, ‘This is the result, now how do you get the original back?’ because you don’t know how many colors went in or the proportion of each.”
It’s the same for chatter in a crowded room.
“You can pull the noise away, but the speech that you’re left with can’t be understood because it’s distorted,” Healy said. “This algorithm pulls the noise away, but the remaining speech is intelligible.”

Read the rest of the story here: http://www.dispatch.com/content/stories/science/2013/12/22/1-cocktail-party-cacophony.html