Incorporated speech overlapped factor (Φ) into speech clarity index (Ψ): method to improve dysarthric speech severity evaluation

Dysarthria is a name given to a group of speech disorders. Our research focuses on developing an automated system for dysarthric speech assessment. Being different from traditional speech assessments which normally rely on human-perceptual analyses, this paper proposes an indicator called modified speech clarity index (ψ^†). By incorporating a speech overlapped factor (φ) into the previous version of the speech clarity index (ψ), the ψ^† achieves a better performance as a speech severity index and speech recognition rate predictor for people with dysarthria. A number of experiments are made to compare predicted recognition rates, generated by ψ^†, with the recognition rates from HMM and ANN system. The effectiveness of the indicator is evaluated in terms of (i) root-mean-square distance, (ii) correlation coefficient, and (iii) rank-order inconsistency. The experiments on a control set and an unknown set of eight dysarthric speakers show that ψ^† achieves the outstanding results when compared to the standard assessments.