May 2016 PRNewswire
Wearers of cochlear implants and hearing aids often have difficulty teasing out what someone is saying over "babble" — the cacophony of other talkers — and other ambient sounds. New York University researchers have devised a novel solution: an algorithmic approach that, like making drinkable water from pond water, distils the talker's voice from a turbid wash of noise.
Most algorithms for acoustic noise suppression aim to eliminate steady background noise — the sound of an air conditioner or road noise are familiar examples — which is relatively easy to attenuate. Babble is much more difficult to suppress because it resembles the foreground voice signal one aims to hear. Few algorithms even attempt to eliminate it. To tackle the problem, Roozbeh Soleymani, an electrical engineering doctoral student, created an innovative noise reduction technology called SEDA (for Speech Enhancement using Decomposition Approach) with Professors Ivan Selesnick and David Landsberger.
The traditional method to analyse a speech signal decomposes the signal into distinct frequency bands, like a prism that separates sunlight into a rainbow of colours. SEDA, however, decomposes a speech signal into waveforms that differ not just in frequency (the number of oscillations per second) but also in how many oscillations each wave contains. "Some waveforms in the SEDA process comprise many oscillations while other comprise just one, waveforms with few oscillations are less sensitive to babble, and SEDA is based on this underlying principle," said Soleymani. Selesnick added that this powerful signal analysis method is practical only now because of the computational power available in electronic devices today.
The potential uses for SEDA, for which a U.S. patent application has been submitted, go way beyond helping the hearing impaired. While it was originally conceived for improving performance with cochlear implants the market might even be bigger for normal hearing people and for mobile phones.