Abstract
Vowel identification in noise using consonant-vowel-consonant (CVC) logatomes was used to investigate a possible interplay of speech information from different frequency regions. It was hypothesized that the periodicity conveyed by the temporal envelope of a high frequency stimulus can enhance the use of the information carried by auditory channels in the low-frequency region that share the same periodicity. It was further hypothesized that this acts as a strobe-like mechanism and would increase the signal-to-noise ratio for the voiced parts of the CVCs. In a first experiment, different high-frequency cues were provided to test this hypothesis, whereas a second experiment examined more closely the role of amplitude modulations and intact phase information within the high-frequency region (4–8 kHz). CVCs were either natural or vocoded speech (both limited to a low-pass cutoff-frequency of 2.5 kHz) and were presented in stationary 3-kHz low-pass filtered masking noise. The experimental results did not support the hypothesized use of periodicity information for aiding low-frequency perception.
Highlights
Speech signals in general cover a wide range of frequencies and usually information across a wide frequency range is grouped to form a single auditory object [1]
The current study examined whether F0-related temporal envelope information derived from high-frequency (4–8 kHz) channels can facilitate the identification of vowels in a masking noise in the low-frequency region below 2.5 kHz
Panel b) shows identification rates that were obtained with low-frequency speech (LFS) and the various high-frequency cues indicated by the gray filled symbols
Summary
Speech signals in general cover a wide range of frequencies and usually information across a wide frequency range is grouped to form a single auditory object [1]. In everyday life speech is rarely perceived in quiet, but in a masking noise and not all parts of the spectrotemporal representation of the speech signal can contribute to speech perception. According to speech perception models such as the speech intelligibility index [2] or the glimpsing model [3], those parts of the representation that have large positive signal-to-noise ratios (SNRs) are most useful for speech perception. Any mechanism that increases the SNR can generally be assumed to improve the perception of masked speech stimuli. Such mechanisms can be external (e.g., a directional microphone in a mobile device or hearing aid) or internal in the auditory system, e.g. selection of appropriate auditory channels that carry specific speech cues. The current study aims at clarifying whether stimulus information derived
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
