Speech perception with cochlear implants and tactile aids

Abstract

Three related areas in our research are speech perception in noise, consonant recognition, and multimodal speech perception. Perception of speech in four‐speaker babble was evaluated with a multiple channel cochlear implant presenting first‐ and second‐formant information. In quiet, scores were 80% for vowels, 48% for consonants, and 51% for sentences. The signal‐to‐noise ratios required to reduce scores to 75% of the value in quiet were 8 dB for vowels, 12 dB for consonants, and 14 dB for sentences. Detailed consonant studies showed that amplitude information contributed directly to the recognition of voicing and manner of articulation and influenced the perception of other cues. In addition to auditory cues, lipreading is commonly used with implants and other aids. A probabilistic model of feature recognition describes the combination of auditory plus visual information well. The combined effect of tactile information is overestimated by the model for minimally trained subjects. The model may clarify issues in the training of hearing‐impaired children, as well as the design of speech processing aids.Three related areas in our research are speech perception in noise, consonant recognition, and multimodal speech perception. Perception of speech in four‐speaker babble was evaluated with a multiple channel cochlear implant presenting first‐ and second‐formant information. In quiet, scores were 80% for vowels, 48% for consonants, and 51% for sentences. The signal‐to‐noise ratios required to reduce scores to 75% of the value in quiet were 8 dB for vowels, 12 dB for consonants, and 14 dB for sentences. Detailed consonant studies showed that amplitude information contributed directly to the recognition of voicing and manner of articulation and influenced the perception of other cues. In addition to auditory cues, lipreading is commonly used with implants and other aids. A probabilistic model of feature recognition describes the combination of auditory plus visual information well. The combined effect of tactile information is overestimated by the model for minimally trained subjects. The model may clarify is...