Date on Master's Thesis/Doctoral Dissertation

12-2012

Document Type

Doctoral Dissertation

Degree Name

Ph. D.

Department

Psychological and Brain Sciences

Committee Chair

Zahorik, Pavel

Author's Keywords

Speech intelligibility; room acoustics; reverberation; perceptual learning

Subject

Speech; Intelligibility of; Sound--Reverberation

Abstract

Speech intelligibility has been found to improve with prior exposure to a reverberant room environment. It is believed that perceptual mechanisms help maintain accurate speech perception under these adverse conditions. Potential factors underlying this speech enhancement effect were examined in three experiments. Experiment 1 studied the time course of speech intelligibility enhancement in multiple room environments. Carrier phrases of varying lengths were used to measure changes in speech intelligibility over time. Results showed an effect of speech enhancement with a time course that varied with the signal-to-noise ratio between the speech and a broad-band noise masker. Additionally, greater speech enhancement was found for reverberant environments compared to anechoic space, which suggests that a de-reverberation mechanism in the auditory system may enhance the temporal processing of speech. Experiment 2 examined the influence of the specific source and listener position within the room environment on speech enhancement. Source and listener configurations in three virtual room environments were altered to create a disparity between the position of a carrier phrase and a following speech target. Results showed robust effects of speech enhancement when the source and listener configuration were mismatched which suggests that speech enhancement relies on the general decay pattern of the room environment and not the specific temporal/spatial configuration of early reflections. Experiment 3 assessed the relationships between room-associated speech enhancement and single-reflection echo suppression by measuring echo thresholds for both a traditional click-based stimuli and with speech materials. Echo thresholds were found to be uncorrelated with the results of Experiment I. This suggests that early reflections have little impact on the de-reverberation aspect of speech enhancement, which is consistent with the results from Experiment II. A two-process hypothesis is proposed to account for the results of these experiments as well as previous research on this topic. Prior exposure to a speech pattern provided via carrier phrases is argued to elicit improved temporal processing of speech that results in speech enhancement. It is also argued that a process of de-reverberation effectively reduces the attenuation of temporal information in room environments.

Download

Share

COinS