AUDITORY AND PHONETIC CODING OF SPEECH

State University Of New York At Buffalo, Buffalo NY

Investigators

Linked publications & trials

Abstract

In language processing, our impression is that speech is a sequence of words. This subjective impression conceals a complex coding process that transforms sound into a segmental representation from which word recognition can proceed. The goal of this research is a formal specification of the auditory and phonetic coding of speech. This, in turn, can serve as a model of normal perception for comparison with theories of perceptual disorders and perceptual development. This research may also contribute to the development of machine-based speech perception which could serve as an aid to the deaf. The experimental studies and computer modeling in this research are aimed at answering three interrelated questions. How are the acoustic qualities of speech mapped onto a segmental representation of the sound? What is the precise nature of this segmental representation? How is the auditory and phonetic coding of this segmental representation influenced by higher level processes, including the listener's knowledge of the words of the language? The approach to answering these questions involves converging operations. Different experimental tasks that tap into different aspects of human perception are used to provide a pattern of data that can distinguish among competing theories. The experimental procedures of identification, discrimination, monitoring, naming, lexical decision, and similarity judgment will be used in conjunction with natural speech, synthetic speech and complex nonspeech stimuli. The choice of stimuli is based on models of the auditory and phonetic coding of speech and models of auditory word recognition. These models allow us to manipulate certain qualities of the speech sounds (the shape of the spectrum of a vowel, the number of words a syllable is similar to) while holding other aspects of the sounds constant. The experimental tasks include both on-line measures related to fluent language processing as well as more traditional, psychophysical tasks. The experimental results will be incorporated into computer simulation models of human speech perception designed to assess the adequacy of the underlying theories. The simulation models can also serve as a basis for understanding perceptual disorders and possible methods for rehabilitation.