Repetition detection in dysarthric speech

Repetition detection is an important pre-processing step in application such as speech to text alignment, voice based interactive system etc. It is very challenging to detect the repeated words because a speaker may utter the repeated words partially or may miss some words in between as it is more often, in the case of Dysarthric utterances. To address these issues, we propose an approach for repetition detection and tested on Dysarthric utterances by extracting features such as MFCC and formants. For calculating similarity scores between the words, we employed two approaches: Dynamic time warping (DTW) and polynomial curve fitting (PCF). Finally, we compared the results of both the approaches by taking each feature independently. DTW based approach found to be more accurate exemplified by the experimental results.