The overall goal of these studies is improvement of communication by hearing-impaired listeners in adverse listening conditions. The long-term objective is to determine factors underlying speech intelligibility, especially in noise and reverberation. For many hearing-impaired listeners who are able to communicate in quiet listening conditions, verbal communication in noisy and reverberant rooms is very difficult. Much of the past effort in facilitating such communication has been directed toward: 1) delivering speech with reduced background noise and reverberation (assistive listening systems), 2) improving speech to noise ratios by selective amplification, and 3) enhancing speech cues. Further advances in selective amplification and speech enhancement techniques require better knowledge about the salience of acoustic cues in degraded listening conditions for hearing-impaired listeners. The proposed studies will focus on cues to vowel and consonant perception in undegraded and degraded listening conditions and specifically: 1) identification of naturally produced vowels by various talkers, 2) vulnerability of specific acoustic cues in computer generated vowel-like stimuli to degradation by either noise or reverberation, and 3) assessment of hierarchy of acoustic cues to voicing distinction in selected consonants, in undegraded and degraded conditions. In the first group of experiments, errors will be related to vowel parameters in individual productions. The understanding of factors underlying speech intelligibility of normal talkers has importance for audiological testing and rehabilitation. In the second group of experiments, vowel continua will be constructed. In each continuum, only one parameter will be a variable and the perceptual boundaries between vowel categories will be established. These boundaries will relate to the salience of the tested variable to degradations by either noise or reverberation. Information about the salience of acoustic cues is important for speech enhancement algorithms and for the formulation of speech perception models in degraded listening conditions. In the third group of experiments involving consonants, effectiveness of cue enhancement in degraded listening conditions will be assessed. Obtained data will also be useful in production of synthesized speech. As computer generated messages become more ubiquitous in listening environments, neither ideal listening conditions nor ideal listeners can be guaranteed. The provision of highly intelligible messages is indicated not only for listener comfort, but also, more importantly, for listener safety and welfare purposes.
