Abstract
Temporal envelope (TE) cues provide a great deal of speech information. This paper explores how spectral subtraction and dynamic-range compression gain modifications affect TE fluctuations for parallel and series configurations. In parallel processing, algorithms compute gains based on the same input signal, and the gains in dB are summed. In series processing, output from the first algorithm forms the input to the second algorithm. Acoustic measurements show that the parallel arrangement produces more gain fluctuations, introducing more changes to the TE than the series configurations. Intelligibility tests for normal-hearing (NH) and hearing-impaired (HI) listeners show (1) parallel processing gives significantly poorer speech understanding than an unprocessed (UNP) signal and the series arrangement and (2) series processing and UNP yield similar results. Speech quality tests show that UNP is preferred to both parallel and series arrangements, although spectral subtraction is the most preferred. No significant differences exist in sound quality between the series and parallel arrangements, or between the NH group and the HI group. These results indicate that gain modifications affect intelligibility and sound quality differently. Listeners appear to have a higher tolerance for gain modifications with regard to intelligibility, while judgments for sound quality appear to be more affected by smaller amounts of gain modification.
Highlights
Many of the 31.5 million Americans with hearing loss are candidates for hearing aids [1]
The acoustic analysis shows that, out of the processing conditions with gain modifications, series provided the least amount of gain fluctuation while maintaining audibility
The natural speech fluctuations in the temporal envelope were more dramatically altered in the compression and parallel conditions, because of the increased amount of gain fluctuations over time
Summary
Many of the 31.5 million Americans with hearing loss are candidates for hearing aids [1]. Results of Kochkin [1, 4] indicate that much of the potential market for hearing aids is not being reached, and that those who are users of hearing aids are not completely satisfied with their instruments. Numerous factors contribute to the lack of satisfaction, some related to speech intelligibility and sound quality. Four others are related to speech intelligibility, including “overall benefit,” “ability to hear in small groups,” “one-on-one conversation,” and “listening to TV.”. Many of these factors relate to listening in the presence of background noise. Hearing aid signal processing algorithms attempt to improve intelligibility and/ or quality in noisy speech by reducing the amount of noise or increasing the audibility of low level speech sounds through modifications to the temporal envelope. Listeners fail to benefit from these signal processing algorithms when compared to simple linear processing, where the only signal modification is amplification (e.g., [2, 5,6,7,8])
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
