An Approach for Objective Assessment of Stuttered Speech Using MFCC Features

Syllable repetition is one of the important parameter in assessing the stuttered speech objectively. The existing method which uses artificial neural network (ANN) and Hidden Markov Model (HMM) requires high levels of agreement as prerequisite before attempting to train and test to separate fluent and nonfluent. We propose automatic detection method for syllable repetition in read speech for objective assessment of stuttered disfluencies which uses a new approach and has four stages comprising of segmentation, feature extraction, score matching and decision logic: Segmentation is assisted manually which is tedious but straightforward. Feature extraction is implemented using well known Mel frequency Cepstra coefficient (MFCC). Score matching is done using Dynamic Time Warping (DTW) between the syllables. The Decision logic is implemented by Support Vector Machine (SVM) and compared with our previous work which uses Perceptron method. The proposed objective approach has an advantage over the manual (subjective), which provide consistent measurement required for assessment. The assessments by human judges on the read speech of 15 adults who stutter are described. 80% of data are used for training and 20% for testing. The average result was found to be 93.45%, which is better than our previous work [80.78%] using HMM.