Investigation of Speech Sounds with an Electronic-Analog Ear

Abstract

A good deal of attention has been devoted to the analysis of speech sounds into intensity-frequency-time patterns. The present investigation is concerned with the related problem of analysis of speech sounds in terms of movement of the basilar membrane. Sounds have been characterized in terms of patterns of velocity of movement at 36 locations along the “basilar membrane” of an electronic-analog ear [W. F. Caldwell, E. Glaesser, and J. L. Stewart, “Design of an Analog Ear,” in Biological Prototypes and Synthetic Systems, edited by E. E. Bernard and M. R. Kare (Plenum Press, New York, 1962)]. Following conversion into digital form, the response of the analog ear was stored in a PDP-1 computer. For visual analysis, the digitized response was displayed on a 16-in. oscilloscope in three-dimensional patterns (location-velocity-time). Initial studies with the analog ear had two main goals: (1) specification of velocity patterns for the various sounds present in speech, and (2) simplification of the velocity patterns by determining the minimum number of basilar-membrane locations required to discriminate the patterns. The second goal has been pursued on the assumption that the locations that have the greatest variability (σ2) as determined across speech sounds would be most important for the discrimination of those sounds.