Research on Vocal Tic Symptom Detecting Using SVM/HMM

摘要

Vocal tic disorder represents a disease of making weird sounds without any reason. This disease must be found and cured in the early stage of the disease in order to prevent development to Tourette’s syndrome. In this paper, we introduce a machine learning based speech recognitiontechnology to detect vocal tic disorder in the early stage. Vocal tic disorder has repetition but an irregular tic pitch that vary in each patient. To identify vocal tic disorder with these characteristics we used MFCC, a speech feature extraction method, and to generate a recognition modelwe used machine learning algorithms SVM and HMM. We confirmed an 88% recognition in the SVM algorithm and a 62% recognition in the HMM algorithm. This confirms that in identifying vocal tic disorders with repetition but irregular tic pitches SVM has a better recognition rate than HMM.