A review of the biological, psychophysical, and speech processing principles used to design the tickle talker

Abstract

The Tickle Talker is a wearable electrotactile speech processor, designed to be used by profoundly hearing-impaired children and adults in conjunction with lipreading and residual hearing. The effectiveness of such a device is affected by an interaction between biological, human engineering, psychophysical, and speech processing considerations. The requirements, the design principles, and the performance of the Tickle Talker in each of these areas will be discussed. Electrical stimulation of the nerve bundles lying along the sides of the fingers was chosen to provide safe, comfortable, energy-efficient stimulation of a well-organised and sensitive part of the tactile sensory system. This is achieved at a small cost to the appearance and mobility of one hand when using the Tickle Talker. The biphasic pulse waveform used to stimulate the nerve bundles has been chosen to ensure a biologically safe stimulus. The electrical parameters (pulse duration, pulse rate, and electrode position) that are used to encode speech information are varied within ranges that are matched to the characteristics of the tactile sense. The usable ranges and information-carrying potential of each of these parameters have been assessed in psychophysical experiments. A comparison of these results with similar experimental data for cochlear implant and hearing aid users is instructive in assessing the possible limitations of tactile and auditory speech processors. The results discussed will include the discrimination and identification of stimuli differing in intensity, duration and pulse rate; the identification of different spatial patterns of stimulation, and the detection of gaps in stimuli.

In most respects, the tactile results are similar to the corresponding auditory measures. The resolution of temporal differences such as pulse rate discrimination or gap detection are generally not as good as in the auditory case, but may be as good or better than the corresponding results for some profoundly hearing-impaired individuals. The speech processor used in the Tickle Talker is a "feature extraction" device that explicitly estimates the second formant frequency, amplitude envelope, and fundamental frequency of the voice and encodes them in terms of electrode position, pulse width and pulse rate of the electrical stimulation pattern. Consideration of the psychophysical results and the speech information available from these parameters allows optimization of the Tickle Talker's operation and a broad estimation of its potential performance in speech discrimination. The perception of duration and place of articulation (front/back) of vowels, and the manner and voicing of consonants are expected to be improved by the Tickle Talker. Prosodic variations conveyed by pulse rate are expected to be perceived by some users, but not all. High frequency consonants such as: /s/,/z/./?/, and /t?/ are encoded in a particularly salient manner by the Tickle Talker.