Acoustic simulation of cochlear implant hearing: Effect of manipulating various acoustic parameters on intelligibility of speech. (2nd January 2018)

• Record Type: Journal Article
• Authors: Jain, Saransh; Vipin Ghosh, P.G.
• Is Part Of: Cochlear implants international. Volume 19:Number 1(2018)
• Page Start: 46
• Page End: 53
• Publication Date: 2018-01-02
• Subjects: Speech perception -- Vocoded speech -- Simulated hearing -- Speech processing -- Signal processing -- Principal component analysis -- PLS; Cochlear implants -- Periodicals; 617.882
• DOI: 10.1080/14670100.2017.1386384
• Languages: English
• ISSNs: 1467-0100

Abstract

Abstract : Objective: Cochlear implants process the acoustic speech signal and convert it into electrical impulses. During this processing, many parameters contribute to speech perception. The available literature reviewed the effect of manipulating one or two such parameters on speech intelligibility, but multiple parameters are seldom manipulated. Method: Acoustic parameters, including pulse rate, number of channels, 'n of m', number of electrodes, and channel spacing, were manipulated in acoustic simulations of cochlear implant hearing and 90 different combinations were created. Speech intelligibility at sentence level was measured using subjective and objective tests. Results: Principal component analysis was employed to select only those components with maximum factor loading, thus reducing the number of components to a reasonable limit. Perceptual speech intelligibility was maximum for signal processing manipulation with respect to 'n of m' and rate of pulses/sec. Regression analysis revealed that lower rate (=500 pps/ch) and lesser stimulating electrodes per cycle (2–4) contributed maximally for speech intelligibility. Perceptual estimate of speech quality (PESQ) and composite measures of spectral weights and likelihood ratio correlated with subjective speech intelligibility scores. Discussion: The findings are consistent with the literature review, indicating that lesser stimulated channel per cycle reduces electrode interaction and hence improve spectral resolution of speech. Reduced rate of pulses/second enhances temporal resolution of speech. Thus, these two components contribute significantly to speech intelligibility. Conclusion: Pulse rate/channel and 'n of m' contribute maximally to speech intelligibility, at least in simulations of electric hearing.