Intelligible Speech Research

TERC / Bridge Multimedia:

Communication Intelligibility: Intelligible Speech

Overview:

Phase I for creating the English and Spanish human voice narrations for the Signing Science and Math Dictionaries, involved cross-disciplinary research to examine best practices for intelligible speech for spoken word recording, an area not previously explored. Populations potentially benefitting include hard of hearing, blind, Deaf-Blind, ASD and English Language Learners. Areas of investigation included:

• Speech/language pathology
• Speech recognition technology
• Speech-supported sign language
• Cognition and language processing
• Spoken word rules for mathematics terms
• Psychoacoustics
• Audio acoustics
• Audio engineering technology

Bridge and TERC contacted experts in speech recognition software, psychoacoustics, and language processing, as well as speech/language pathologists to identify clear speech methods to incorporate into the narrations to increase speech intelligibility—in both English and Spanish—for people with auditory impairments. The methods identified included adjusting the rate of words following each other; use of pauses within sentences and use of longer pauses than typical where there are commas and periods; attention to the articulation of sounds within words of longer duration; accuracy of word pronunciation; increasing "consonant power" of letters—such as T and P; M and N—that are difficult to differentiate; attention to pitch contour and consistency of word intonation, and selection of individual word pronunciations based on ease of recognition.

Another component of this speech intelligibility study was to integrate the "spoken word rules" for mathematics in standard use by blind/low vision students. The spoken word guidelines were expanded to accommodate science terms that had previously not been included in the protocols.

Until now, clinicians have used these methods, but they have never been applied as an approach to recording human voice for audio books, instructional materials, dictionaries, etc. As such, the integration of this "intelligible speech" component into materials for the intended audience is in itself groundbreaking.

Full report coming soon.