Intelligibility Test Methods and Procedures for the Evaluation of Speech Communication Systems

摘要

In further exploring the Modified Rhyme Test (MRT), a recently developed intelligibility test designed for the evaluation of speech communication systems under operational military conditions, research has been conducted in the following areas: (a) the relation between MRT scores and other intelligibility test scores for various types and levels of speech distortion; (b) the influence of the closed-response format and listening experience on MRT scores; and (c) speaker intelligibility and the selection of speakers for recording the test lists. The present report describes the work undertaken in each of these areas. The ultimate objective of the work is the development of valid procedures for the efficient evaluation of speech communication systems. The major experimental results demonstrate that (1) the relation between scores obtained with different intelligibility test materials is not unique but depends considerably on the type of speech distortion employed, (2) neither the closed-response format nor prior listening experience appreciably affects MRT scores, and (3) less intelligible speakers tend to be those whose voiceless consonants are generated with lower intensity, particularly in word-final position. (Author)