A Comparison of Two Training Strategies for Speech Recognition with an Electrotactile Speech Processor

Abstract

Seven normally hearing adults were trained in the use of a multiple-channel electrotactile speech processor for 70 hours over a 6-month period. Two training strategies were used on each of the subjects: (1) analytic-plus-synthetic (AS), and (2) synthetic (S). The speech perception abilities of the subjects were assessed with closed sets of vowels and consonants, open sets of words and sentences, and speech tracking, for the tactile-plus-lipreading, tactile, and lipreading conditions. The subjects were tested on three separate occasions: (1) at the beginning of the study, (2) after 35 hours of training, and (3) after a further 35 hours of training. Wilcoxon matched-pair signed-ranks (MPSR) tests showed that improvements observed with both the AS and S strategies were significantly (p < .05) greater than zero for most tests and conditions. The Wilcoxon MPSR test showed that the difference in improvements between the training strategies was significant for only the vowel and consonant tests in the tactile condition.