Characterizing early markers of degraded speech encoding: temporal fine structure and envelope cues

Speech contains rapidly varying temporal fine-structure (TFS) information as well as slower temporal envelopes (TENV). Both features are linked to speech intelligibility, although the presence of TFS is thought to boost speech reception in adverse listening conditions. Since the respective role of cochlear synaptopathy and outer-hair-cell (OHC) deficits to degraded speech reception is unclear, it is important to clarify which ENV and TFS information is available at the level of the auditory midbrain in the normal and impaired auditory system. Starting from the earliest neuronal correlate of an speech, we aim to understand how TFS and TENV cues are represented in a computational model of the (impaired) human auditory periphery (Verhulst, Altoè, & Vasilkov, 2018). The outcome of this study can guide the development of hearing restoration strategies tailored to listeners with synaptopathy and/or OHC loss.

Speech was decomposed into TFS or TENV chimaera and stimuli were either band-pass, high-pass or low-pass filtered to target different aspects of cochlear sound encoding. An additional TFS stimulus, based on the zero crossings of the fundamental waveform and the first harmonic, was added to prevent a reconstruction of TENV cues in the signal decomposition. Models with different combinations of synaptopathy and outer-hair-cell deficits were considered. To study which aspects of the stimuli are preserved after (impaired) cochlear processing, cross-correlations were performed between the stimuli and stimulated population auditory-nerve (AN) responses. Additionally, the relationship between simulated AN responses to the unmodified and (filtered) modified stimuli was investigated to clarify the respective role of TFS/TENV and cochlear frequency regions to speech representation in the (impaired) auditory system.