Abstract:

Humans listening to speech in a small room are frequently unaware of reverberation. It is unknown if neurological processes remove these echoes or if they are simply disregarded when speech is phonetically processed. In other words, is there a neurological mechanism that is capable of removing echoes to create a clean speech neurological signal before phonetic processing? Or is the brain capable of processing reverberant phonemes? Word intelligibility experiments examine these questions. Preliminary experiments investigate how characteristics of simulated reverberation such as room size, absorption of walls, source location, and listener position affect intelligibility. The results of these experiments are used in the design of primary experiments that address human capabilities. In the capability tests the effects of binaural listening, reverberation level, and deconvolution processing are investigated. These experiments approach the fundamental questions through the use of a three-factor experiment (the factors being binaural versus diotic, high versus low levels of reverberation, and simulated reverberation versus convolutional noise). Through the primary and interaction effects of these factors the data illustrates the extent of neurological dereverberation.