Improved Speech Recognition for High-Frequency Hearing Loss

Back to all technologies
Download as PDF
With conventional hearing aid processing, individuals who have severe to profound mid- to high-frequency hearing loss still have great difficulty perceiving important speech information. Alternative processing strategies that move inaudible high-frequency information to lower frequency regions where hearing loss is less severe have been proposed; however, they all have a common flaw in that they tend to increase confusion between certain sounds, especially 's' and 'z'. The seriousness of this problem is illustrated by the fact that these sounds constitute about 8 percent of all spoken consonants in the English language.

Purdue University researchers have developed a unique digital signal processing technique that recodes high-frequency speech information in a way that enhances perceptual differences between commonly confused speech sounds. In preliminary testing, the algorithm has worked better than commercially available alternatives, especially for the consonants 's', 'sh', 'ch', 'j', 't', 'k', and 'z'. No degradation in vowel recognition has been observed. The efficacy of using this technology for a broader range of hearing losses, including those with cochlear implants, is currently being tested.

-Improved identification of consonants
-Vowel degradation has not occurred
-Preliminary testing shows improvement over current commercial alternatives

Potential Applications:
-Heath Loss Technology
Sep 1, 2015
United States
Sep 25, 2018

May 30, 2013
Utility Patent
United States
Oct 27, 2015

May 31, 2012
United States
Purdue Office of Technology Commercialization
The Convergence Center
101 Foundry Drive, Suite 2500
West Lafayette, IN 47906

Phone: (765) 588-3475
Fax: (765) 463-3486