期刊名称：International Journal of Computer Science and Network Security
印刷版ISSN：1738-7906
出版年度：2017
卷号：17
期号：7
页码：188-195
出版社：International Journal of Computer Science and Network Security
摘要：Spoken utterances play significant role in identifying the emotional states of speakers. However, extracted features add sense in spoken utterances which leads to provide speaker emotions. In this paper, a novel approach is presented for feature extraction toward classifying normal and special children speech emotion using spoken utterances in Urdu language. Eleven different features presents in this paper using thresholding technique on the extracted features implements the proposed algorithm namely: frequency, pitch, rate of zero passages, rate of acceleration, formant frequencies, intensity, log power, log energy, Mel Frequency Cepstrum Coefficients (MFCC), Linear Prediction Cepstral coefficient (LPCC) and Relative Spectral Transform - Perceptual Linear Prediction (Rasta PLP) with four different emotions (Angry, Happy, Neutral and Sad) to classify speech emotion of normal and special children. Experimental results evident that the proposed algorithm shows 100% accuracy with reduce error rate in Normal speech emotion and special speech emotion category for Angry, Neutral, Sad and Happy and Sad emotions respectively.
关键词：Feature Extraction; Speech Emotion; Normal and special children; Urdu Language; classification accuracy.
