Evaluating the effects of dereverberation techniques on speech understanding in different reverberant environments
Links to Fileshttp://library.towson.edu/digital/collection/etd/id/68957
MetadataShow full item record
Type of Workx, 103 pages
DepartmentTowson University. Department of Audiology, Speech-Language Pathology and Deaf Studies
Reverberation and background noise is present in almost all everyday listening situations. Reverberation times in real rooms vary widely from nearly several milliseconds up to several seconds based on the absorptive properties of the materials present in a room. In a reverberant listening environment, listeners receive speech through three main components: direct sound, early reflections, and late reflections. Older adults with hearing loss are more susceptible to the negative effects late reflections have on speech. On the other hand, previous studies have indicated that early reflections are associated with improved speech understanding. With that being said, it is surprising that even with the availability of de reverberation processing strategies, most hearing aid manufacturers do not implement these processing strategies in their devices. The effects of a dereverberation processing strategy on speech understanding in two reverberant environments (T60 = 1s and T60 = 2s) were evaluated in this experiment. Target-to-Masker (TMR) identification thresholds from the participants in this study were analyzed in terms of 3 different conditions. These conditions include: three different reverberation times (i.e., T60 = 0s, T60 = 1s, and T60 = 2s); two different spatial configurations (colocated and spatially separated), and three different dereverberation processing techniques (i.e., correct, over, and underestimation). The results of experiment 1 demonstrated that the listener's ability to identify the target call-sign in a multi-talker environment improved with spatially separating the target speaker from the masker speakers. Consequently, as reverberation increased speech understanding was significantly reduced. In experiment 2, the underestimation Binaural Room Impulse Response elicited the lowest TMR identification thresholds. The conditions were ordered: Under, Correct, and Overestimation (going from the lowest to the highest TMR). Multiple regression analyses predicting the amount of Spatial Release from Masking using age and PTA indicated that only PTA was significant in predicting SRM in the correct estimation condition in the reverberant condition of 2 seconds. The results of this study provide evidence that adding a dereverberation program utilizing an underestimation processing strategy as an option for hearing aid users could improve speech understanding and reduce listening effort in reverberant environments.