Improving cochlear-implant speech perception in noise

With the state-of-art technology, most cochlear-implant users derive substantial benefit in understanding speech in quiet listening conditions. However, their ability to understand speech deteriorates significantly in background noise. One possible reason behind implant users' poor speech understanding in noise is the lack of temporal fine structure in cochlear-implant processing. Here, we implemented and evaluated two noise suppression techniques to improve speech perception of cochlear-implant users in noise. The first noise suppression technique studied is called 'Companding' which mimics the two-tone suppression phenomenon in a normal ear to produce spectral enhancement. The key parameters were identified and investigated to optimize the companding performance. Cochlear-implant users tested on speech perception tasks over a wide range of signal-to-noise ratios showed significant benefit with the companding strategy. The second noise suppression strategy combined spectral and temporal enhancements by expanding the narrow-band envelopes in the input signal and additionally processing with 'Transient Emphasis Spectral Maxima' (TESM) strategy to enhance the weak short-duration cues important for consonant perception. Cochlear implant users showed significant improvement in understanding speech in noise with both expansion and expansion combined with TESM. Although these techniques improved the speech understanding of implant users in noise significantly, their performance was far from satisfactory. This led to the final study which investigated the role of temporal fine structure in speech perception. The technical issues related to the extraction of fine structure and the relative contributions of temporal envelope and fine structure were studied by testing normal-hearing listeners. This study delineated the differences between temporal envelope and temporal fine structure both from methodological and perceptual viewpoints. Together, the present work suggests that bio-mimic processing can improve cochlear-implant performance.