Efficacy and viability of an algorithm to improve speech understanding in noise for the hearing impaired
Eric W. Healy · The Journal of the Acoustical Society of America · 2016
A primary complaint of hearing-impaired individuals involves poor speech understanding when background noise is present. Hearing aids and cochlear implants often allow good speech understanding in quiet backgrounds. However, the listeners’ noise intolerance and the devices’ inability to effectively combat background noise often conspire to produce poor performance in noise. Perhaps surprisingly, effective solutions to this problem have remained elusive despite considerable effort. One promising solution involves a single-microphone algorithm to extract speech from background noise. The algorithm is based on the concept of the ideal binary or ratio mask, and employs standard machine-learning techniques to train a deep neural network to estimate the mask, given only the speech-plus-noise mixture. Existing data indicate that large intelligibility increases by hearing-impaired listeners may be obtained across a variety of noisy conditions. In this talk, an overview of this approach will be provided, and the potential for implementation into hearing aids and cochlear implants will be discussed. [Work supported by NIH.]