ASA 129th Meeting - Washington, DC - 1995 May 30 .. Jun 06

4aPP7. Prediction of articulation scores.

Christine M. Rankovic

Dept. Speech-Language Pathol. and Audiol., Northeastern Univ., 360 Huntington Ave., Boston, MA 02115

Harvey Fletcher developed articulation theory to enable the prediction of the speech transmission capacity of imperfect communication systems, thereby eliminating the need for the time-consuming speech recognition testing that is otherwise required to differentiate among systems. Two assumptions underlie the theory: (1) a speech signal in each of a number of contiguous frequency bands comprising the audible spectrum contributes independently to the articulation score (percent-correct for lists of nonsense syllables); and (2) the separate contributions are additive. The approach of Fletcher and his colleagues was to define the articulation index, A, an entity for which these assumptions hold true. The equations for calculating A contain variables representing hearing acuity, critical bandwidth, masking, loudness, perception, and frequency importance weights so that the communication system's frequency response and gain are translated into perceptual terms. In addition to the derivation of the theory, Fletcher's 1953 book provides a method for calculating A using charts and tabled data, as well as an extensive validation study that included a variety of suboptimal systems. Fletcher's presentation of articulation theory is far more comprehensive than two popular versions [ANSI S3.5-1969; N. R. French and J. C. Steinberg, 90--119 (1947)]. Differences among these versions of the A calculation and consequences for predicting articulation scores will be discussed. [Work supported by NIH.]