Roy D. PattersonProfessor

We are trying to understand how the ear and the brain construct the perceptions, or auditory images, that we hear when presented with the sounds of speech and music. When children and adults say the same words, the acoustic waves arriving at the ear are very different because of the difference in speaker size. Nevertheless, we hear the same speech message (the words). Speech recognition machines have great difficulty with this problem. The data from human listening experiments suggest that the ear and brain construct an internal representation of the speech that is “scale invariant,” that is, the same for people of all sizes. With colleagues in Japan, we have developed a mathematical transform to show how the cochlea and auditory neurons could make such a representation, and how the representation could assist in attempts to produce better speech recognition machines. With colleagues in the UK and Germany, we perform brain imaging experiments to establish where the different stages of the processing take place in the brain.