Hybrid Approach to Detect Prolonged Speech Segments

In the last 10 decades various methods have been introduced to detect prolonged speech segments automatically for stuttered signals. However less attention has paid by researches in detection of prolongation disorder at parametric level. The aim this study is propose a hybrid approach combining spectral parameters with their recognition accuracies reconstructed signal. paper presents considering individually, parameters, validation system, MFCC feature extraction process, basic model proposed are simulated and experimented on UCLASS derived dataset. Obtained results compared existing works word It observed that yield 92% rate larger frame sizes 200ms when modeled SVM. also tabulated discussed metrics like sensitivity, specificity accuracy detecting segments. focuses characteristics vocalized non-vocalized sounds phoneme 71% Vocalized vowel phonemes over Objectives: objective work algorithm signal disorder. other evaluate obtained performances applying evaluation models compute further extended bring out importance variable size concept analyze variations sounds. Methods: adopted two levels namely preprocessing modeling levels. Preprocessing level an individual level, combing Centroid, Entropy, Energy, ZCR method. A new method applied using Specificity, Sensitivity validate performance. above clustering classification K-means, FCM performance these considered evaluating estimating segment signals All detail following sections. Findings: Hybridizing major finding work. found sensitivity plays role designing validating model. From experiments it identified verification suits better lengths considered. SVM perform all considerations. Novelty: As per Literature survey few prolongation. But not addressed or more than novelty lies selecting stage Spectral centroid entropy as appropriate along Energy parameters. Hence hybridizing automatic system. Novelty brought build system