Speech Feature Recognition with an Electrotactile Speech Processor

The performance of eight subjects was assessed on a closed-set tactual test battery to evaluate elliciency of the speech feature encoding strategy currently used in the University of Melbourne multichannel electrotactile speech processor. The test battery included twelve subtests of suprasegmental and segmental speech feature contrasts. Results showed that all subjects scored significantly above chance on suprasegmentalfeatures such as syllable number, stess and vowel length. In addition, seven of the eight subjects scored significantly above chance for vowel formant frequency discrimination. Scores for manner of articulation contrasts were more variable, with better performance for the higher frequency, longer duration fricatives ana affricates. Scores for voicing contrasts suggested that improvements to the tactual coding of this feature could be achieved. The second study examined the contribution of the tactual input to consonant feature identification for subjects using the electrotactile aid in combination with either a low-pass filtered auditory input or different levels of aided residual hearing and lipreading. Results for four normally-hearing subjects showed increased voicing and manner identification scores with the addition of tactual input. Similar results were found for two hearing-impaired sub­ Jects, one using the tactual input in combination with lipreading and one using the tactual input in combination with aided residual hearing. INTRODUCTION More recently, the combination of visual and tactual input has been shown to be effective in Efforts to utilize the tactile modality as an the coding of both suprasegmental and alternative avenue to convey speech informa­ segmental features of speech. The combined tion for the hearing-impaired date from the early tactual and visual input has also been demon­ work of Gault (1926) with simple vibrotactile strated to improve performance on both closed­ devices. Traditionally, the aim has been to utilize set and to a lesser extent on open-set word and the tactile modality as a sensory substitute for sentence discrimination (Plant, 1986; Brooks et a damaged auditory system. Studies have aI., 1986b; Cholewiak and Sherrick, 1986; demonstrated that prosodic features such as Weisenberger and Miller, 1987. Cowan et aI., syllable length, stress pattern and vowel length, 1988). Successful use of the tactual modality to as well as segmental features such as con­ supplement information available through aided sonant voicing and manner of articulation may residual hearing, or the combination of aided be transmitted through the tactual modality hearing and lipreading has been reported for (Sparks et aI., 1979; Plant, 1983; Saunders and multichannel tactile devices (Eilers et aI., 1988; Franklin, 1985; Boothroyd and Hnath, 1986; Kozma-Spytek and Weisenberger, 1987; Lynch Brooks et a\., 1986a: Carney, 1988; Blarney et et aI., 1988; Cowan et aI., 1989). aI., 1988). Recognition of whole words present­ A wide variety of strategies have been ed tactually through a multichannel vibrotactile utilized to encode speech information as tactile device has also been demonstrated for one patterns and it is reasonable to assume that the extensively trained subject (Brooks, 1987). degree of improvement in speech discrimina­