Reducing Background Noise
A busy office with simultaneous conversations, phones ringing and printers humming can wreak havoc with voice recognition software.
Background noise also makes it hard to be heard during a cell phone conversation.
 Dr. Parham Aarabi, an electrical and computer engineer at the University of Toronto, has developed a system to filter out this intrusive background noise. “You only hear the person of interest,” he says.
The NSERC-funded researcher’s system uses a customized integrated circuit (IC) chip that takes sound from two microphones instead of one.
“We’ve shown that this system gets 30 percent better speech recognition than the next best technique,” he observes.
Working in collaboration with NSERC researcher Dr. Ali Sheikholeslami and Ph.D. students David Halupka and Guangji Shi, Dr. Aarabi has developed a system which measures the phase of the sound waves reaching the two microphones. The system then uses the differences in the arrival times of the signals to target the speaker of interest and reduce the background noise to a faint hum.
This approach allows Dr. Aarabi to build smaller IC chips that are less expensive and less power hungry.
Within two years, he expects to add amplifiers, conditioners and filters onto one chip, thereby obtaining a miniature microphone array system that could be used for handheld computers, cellphones and hearing aids.
Dr. Aarabi is eager to commercialize his system. “I can’t wait to get it out there so people can use it,” he explains.
Contact:
Dr. Parham Aarabi
Department of Electrical and Computer Engineering
University of Toronto
Tel.: (416) 946-7893
E-mail: parham@ecf.utoronto.ca
Web site: http://www.ecf.utoronto.ca/~parham/
|
