Speech Recognition as a Function of Channel Capacity in a Discrete Set of Channels

There have been many attempts to perform a frequency analysis of speech and to use the outputs of this analysis to provide cutaneous stimulation. The results of these tests have left in doubt the issue of whether or not cutaneous recognition of speech is actually possible. In addition to other difficulties, an optimum frequency analysis has never been achieved. Instead, filtering configurations have been chosen essentially arbitrarily. The systems considered may well have had insufficient channel capacity for speech recognition even in the event that the tactile stimulators were arranged in an optimum manner. In this paper, several sets of finite frequency band to discrete channels filters were considered. The frequency of each discrete channel was constrained to the fixed center frequency of the corresponding band so as to be directly translatable to the position of a tactile stimulator. Tests were conducted to measure the auditory recognition rate of speech, resynthesized as the sum of these discrete channel signals. From these measurements with different numbers of channels, a representation of recognition rate as a function channel capacity was obtained. It is anticipated that these results will be helpful in the choice of the optimum filtering configuration for test in actual tactile recognition.