Consonant testing of a tactile hearing aid based on amplitude‐ and time‐quantized speech

Abstract

Speech was first prefiltered with a second‐order bandpass filter centered at 1.5 kHz. Speech was then amplitude quantized to two levels based on the zero crossings of the waveform. This signal was then time quantized such that it could only change states in association with a clock signal. The intelligibility of the processed speech remains high for clock frequencies above 10 kHz (which was the quantizing rate used for this study). The discrete digital frequencies in the resulting waveform were easily detected with a digital logic circuit. These were then displayed onto the skin of the abdomen with a linear array of concentric ring electrodes. Thirty‐one electrodes (representing frequencies between 151 and 2500 Hz) were used for the array. Biphasic constant current pulses were used as the stimuli. Five normal hearing college age students served as the subjects. Discrimination of consonant pairs was used to test the ability of the tactile aid to transmit speech information to the subjects. The following ranges of accuracy (corrected for guessing) were observed: stops and constituents, 18% to 87%; voiced and unvoiced consonants, 3% to 27%; consonants with different places of articulation, 27% to 90%; nasal, glides, and semivowels, 17% to 49%.