|James M. Hillenbrand|
The common, everyday activities of producing and understanding speech are almost never thought of among the skills that we consider to be especially complicated or difficult. The kinds of skills that we usually think of as being complex and challenging are those that require a good deal of explicit practice and instruction - skills like hitting a curve ball, playing the French horn, or even learning to ride a bicycle. Speech is obviously quite different since it is only in special cases that any explicit practice or instruction is needed (these special cases, of course, form much of the foundation of our profession). In spite of this apparent simplicity, the mechanisms that are involved in planning and controlling articulatory movements, and in deriving meaning from a speech signal, are enormously complex and poorly understood. In terms of the number of different structures and muscles that are involved, the speed of movement, and the precision that is needed in both timing and placement, speech production is probably the most complex motor activity that humans perform. On the receiving end, listeners need to cope with a surprising amount of variability in the acoustic structure of the speech signal that results from variation in factors such as individual differences in vocal tract anatomy, dialect, rate of speech, and the phonetic environment in which a particular speech sound occurs. The effect of all this variability is that a given speech sound can be realized with a wide variety of different acoustic signatures. Research in our Speech Acoustics Laboratory has focused mainly on trying to understand the auditory and pattern-matching mechanisms that are involved in recognizing the speech sounds that form the sound pattern of English. Our experiments typically involve acoustic analyses of large databases of speech recorded from a variety of different speakers, developing hypotheses about underlying pattern-matching mechanisms on the basis of those analyses, and then testing those hypotheses with listening experiments that use computer-generated or computer-modified speech signals. Our laboratory is also involved in research that is designed to improve our understanding of the physiological and perceptual mechanisms that are responsible for variations in voice quality.