Assessment of dysarthric speech and an analysis on velopharyngeal incompetence

Abstract

In the present work the role of a speech recognition system in the assessment of dysarthric speech is studied. Initially a continuous speech recognition system is developed for the assessment of dysarthric speech. The performance of the continuous speech recognition system on the assessment of dysarthric speech is found to be dis-satisfactory due to greater number of insertions. Analysis conducted on duration of phoneme and speech rate confirms the fact that the more number of insertions in the recognizer output is due to lower speech rate of dysarthric speakers. To overcome the problem with the continuous speech recognition system an isolated-style speech recognition system is developed. The performance of this system on the assessment is compared with the Frenchay dysarthric assessment (FDA) scores provided with the Nemours speech corpus. From the performance of the isolated-style speech recognition system it is observed that apart from the articulatory problems, some of the speakers are affected with velopharyngeal incompetence also and is analyzed with group delay function based acoustic measure for the detection of hypernasality on dysarthric speech.