Analysing vocal quality of connected speech using Kay's computerized speech lab: A preliminary finding

Abstract

The number of studies that have investigated acoustic correlates of vocal quality using connected speech is limited because the perturbation measures generally require a steady state signal. Kay Elemetrics' Computerized Speech Lab (CSL) and the Multidimensional Voice Program (MDVP) are said to be tolerant to the variation in connected speech signals and may be used for such purposes. This study represents an attempt to investigate whether the analysis is capable of distinguishing dysphonia from normal voice using connected speech. The results showed that only one acoustic measures (relative average perturbation) was sensitive to dysphonia. The limitations of using acoustic parameters to quantify vocal quality in pathological voices may have been attributed to the algorithms used by the specific measurements.