Back     Close

A Comparison of Intelligibility Measurement Techniques for Dysarthric Speech
Jacqueline Laures-Gore & Gary Weismer

Measurement of the intelligibility of dysarthric speakers has primarily taken the form of word identification tests and scaling procedures (Schiavetti, 1992). Word identification tests involve a listener transcribing what has been perceived with percent accuracy of this transcription indicating the intelligibility of the speaker. Scaling procedures take the form of direct magnitude estimation with a modulus, free modulus magnitude estimation, or interval scaling. Weismer and Laures (2002) suggest that magnitude estimates may provide a more complete representation of speech intelligibility than other methods of measurement because it may be more sensitive to non-segmental aspects of speech, such as prosody and voice quality. However, the empirical data supporting such a statement is quite limited. The current study explores the relation between the different types of measurement techniques used for dysarthric speech. Twelve listeners measured the speech of six dysarthric speakers and two healthy control speakers using the four different measurement techniques. Each speaker produced three sentences twice. The sentences were presented via a loudspeaker in a sound attenuated booth. A correlation analysis revealed that the four techniques were highly related. Implications of this finding are discussed.

Back     Close