The effect of spectral enhancement on speech recognition performance of normal-hearing and hearing -impaired individuals
Abstract
The purpose of the present study was to investigate the effect of spectral enhancement on the speech recognition ability of normal-hearing and hearing-impaired individuals. Enhancement of spectral contrast in speech has been proposed as a signal processing strategy in hearing aids to counteract the ill effects of poor frequency resolution in individuals with sensorineural hearing loss. A spectral enhancement algorithm was designed in collaboration with J. Brandon Laflen of Department of Biomedical Engineering at Purdue University to preprocess the speech material. Fifteen subjects with moderate to moderately severe sensory hearing loss and eight subjects with normal hearing were tested at four levels of spectral enhancement using nonsense syllables, vowels, and connected sentences. The four levels of spectral enhancement ranged from 0 to 15 dB of enhancement. In addition to the speech recognition testing, subjective judgment of overall sound quality was evaluated using a paired comparison task. Results indicated that there was no improvement in speech understanding scores with spectral enhancement, rather there was a significant decrease in speech recognition scores with the two high levels of spectral enhancement in the hearing-impaired group. The paired comparison task indicated that while the normal-hearing subjects preferred the overall sound quality of the non-enhancement condition over the spectrally enhanced conditions, the hearing impaired group's preference for the sound quality of the four spectral enhancement conditions were equivocal.
Degree
Ph.D.
Advisors
Novak, Purdue University.
Subject Area
Audiology
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.