Effects of Peripheral Tuning on the Auditory Nerve’s Representation of Speech Envelope and Temporal Fine Structure Cues

Abstract A number of studies have explored how speech envelope and temporal fine structure (TFS) cues contribute to speech perception. Some recent investigations have attempted to process speech signals to remove envelope cues and leave only TFS cues, but the results are confounded by the fact that envelope cues may be partially reconstructed when TFS signals pass through the narrowband filters of the cochlea. To minimize this reconstruction, investigators have utilized large numbers of narrowband filters in their speech processing algorithms and introduced competing envelope cues. However, it has been argued that human peripheral tuning may be two or more times sharper than previously estimated, such that envelope restoration may be stronger than originally thought. In this study, we utilize a computational model of the auditory periphery to investigate how cochlear tuning affects the restoration of envelope cues in auditory nerve responses to “TFS speech.” Both the envelope-normalization algorithm of Lorenzi et al. (Proc Natl Acad Sci USA 103:18866−18869, 2006) and the speech-noise chimaeras of Smith et al. (Nature 416:87–90, 2002) were evaluated. The results for the two processing algorithms indicate that the competing noise envelope of the chimaeras better reduces speech envelope restoration but does not totally eliminate it. Moreover, envelope restoration is greater if the cochlear tuning is adjusted to match Shera and colleagues’ (Proc Natl Acad Sci USA 99:3318−3323, 2002) estimates of human tuning.