The importance of temporal fine structure information in speech at different spectral regions for normal-hearing and hearing-impaired subjects

Abstract

Speech reception thresholds (SRTs) were measured for target and competing-speech signals, processed to contain variable amounts of temporal fine structure (TFS) information. Signals were filtered into 30 1-ERB(N) wide channels (where ERB(N) refers to the bandwidth of normal auditory filters), which were either tone vocoded, preserving temporal envelope information (extracted using the Hilbert transform), or left unprocessed, containing both TFS and envelope information. Improvements in SRT were compared when TFS was progressively introduced, starting with the high- or low-frequency channels. The results suggest some redundancy in the TFS information across frequency regions. In a second experiment, the signal was divided into five, 6-ERB(N)-wide spectral regions, four of which were tone vocoded. The remaining region was either absent (creating a spectral notch) or was present and unprocessed. SRTs were measured for normal-hearing and hearing-impaired subjects. Conditions where all channels were vocoded or unprocessed were also included. Normal-hearing subjects benefited similarly when TFS information was added to each region, suggesting that TFS information is important over a wide frequency range. Hearing-impaired subjects benefited less, although the benefit varied across subjects. Benefit from TFS information in speech was correlated with a psychophysical measure of TFS sensitivity obtained at two center frequencies.