Innovative Monitoring Device for Speech to Challenged Listeners

Period of Performance: 09/01/2009 - 08/31/2011

$176K

Phase 1 SBIR

Recipient Firm

Speech Technology/applied Research Corp.
Bedford, MA 01730
Principal Investigator
Principal Investigator

Abstract

DESCRIPTION (provided by applicant): We will develop a real-time monitoring device for speech intelligibility, to remind or train audiologists and other conversation partners of hearing-impaired listeners. In Phase I, we first will record and evaluate the speech of 30 native English talkers. Building on previous successful findings, we will determine and refine the speech-acoustic parameters that correspond to the articulation of highly intelligible speech, and that distinguish a given talker's less intelligible from more intelligible speech. These parameters will thus relate the listener's result, understanding what was said, to the talker's attention to precise articulation. The talkers will produce less and more intelligible speech samples by using typical and "clear" styles of speech. Clear speech is produced by the instruction, "Speak as if to a hearing-impaired or non-native listener." In nearly all adults, it produces more precisely articulated speech that is also better understood by listeners. Listeners will evaluate the speech produced in both styles by transcribing sentences that they have not previously heard from these talkers. The task will be made more difficult by mixing a high level of noise. This reliably reduces the intelligibility of typical speech much more than of clear speech. Afterward, we will test a simplified, software-only version of the system on clinician students. Such students are routinely trained in clear speech for the future benefit of their clients. Our simplified system will remind them periodically to speak certain randomly selected paragraphs clearly. It will evaluate the result and show the user a score to indicate how close the result came to the earlier talkers'most intelligible versions of that paragraph. The score will be based on the previously derived speech-acoustic parameters. In Phase II, we will extend this to spontaneous and conversational speech. Then we will design a self-contained, handheld device to apply this processing to evaluate a user's articulatory precision, and perhaps flash or vibrate if the precision becomes low. The device will be used to remind or to train the conversation partners of hearing-impaired listeners. Such users will include audiologists and speech pathologists, as well as interested family members. PUBLIC HEALTH RELEVANCE: We will develop a real-time monitor of speech intelligibility. This will be a self-contained, handheld device to evaluate a user's articulatory precision, to remind or to train hearing- impaired listeners'conversation partners, including audiologists and speech pathologists, as well as interested family members. This will enhance communication effectiveness with hearing-impaired listeners and improve quality of life for them, their families, and their communities.