Development of an Automatic Speech Processing Based Procedure for Assessment of Stuttering in Malayalam Speaking Adults through Virtual Mode

Abstract

The current research aimed to develop and evaluate a virtual mode approach using automatic speech processing for assessing fluency parameters in Malayalam-speaking adults who stutter. Objectives of the study were; a. to assess the fluency parameters such as frequency and duration of filled pauses, blocks, repetitions and prolongations, using the stuttering severity instrument – 4th edition (SSI – 4) in virtual mode, b. to assess the fluency parameters using automatic speech processing from the standard Malayalam passage spoken by 15 adults with stuttering (age range of 18-35 years), recorded through Zoom app, c. to compare the values of fluency parameters obtained from automatic speech processing with those derived via perceptual assessment, and d. to analyze the online assessment's efficacy based on feedback from participants and three SLPs. Each participant was asked to read the standard Malayalam passage presented using the Zoom app's presentation mode. The session was recorded and stored on a computer for further analysis. The results of perceptual evaluation and automatic speech processing were compared using Wilcoxon signed rank test. The results revealed no significant difference between the detection of stuttering events through perceptual evaluation and automatic speech processing. The overall accuracy in detection of fluency parameters ranged from 36% to 44%. Also technical and clinical quality of the online assessment was found to be satisfactory by the participants as well as SLPs. The study established the viability of doing online assessments for fluency disorder. Hence the study's findings will promote telepractice for stuttering evaluation, particularly in epidemic situations like COVID-19 where it is difficult to perform a conventional face-to-face examination.