Development Of Speech Recognition System In Social Signal Processing Application

Ngew, Chi Nee (2015) Development Of Speech Recognition System In Social Signal Processing Application. Project Report. UTeM, Melaka, Malaysia. (Submitted)

Text (24 pages) Development Of Speech Recogntion System In Social Signal Processing Application 24 Pages.pdf - Submitted Version Download (318kB)

Abstract

Social Signal Processing (SSP) has been widely used in robotic and computer as one of the Artificial Intelligent (AI) in contribute to human machine interaction. One of the examples of SSP is to recognize human emotions. In this research, a system which capable to recognize different states of emotion in speech is successfully developed using Support Vector Machine (SVM) technique. The first two main objectives of this research are to develop a speech emotion recognition system and graphical user interface (GUI) using MATLAB software. Besides that, performance of the system also has been studied based on the percentage accuracy. Linguistic Data Consortium (LDC) is used as the database. The features contained in the LDC voice samples are extracted and used to develop the dataset. The methods used to extract the features include energy, pitch, formant, Mel-Frequency Cepstrum Coefficient (MFCC) features. Statistic such as mean value is calculated. A training model is introduced to train the classifier in the system and a testing model is used to analyse the system. ‘Happiness’, ‘Anger’, ‘Sadness’ and ‘Neutral’ are examined. Gender dependent and independent test are studied to analyse the gender impact on the performance of emotion recognition. The result shows that the gender independent test has higher accuracy than the gender dependent test. Besides that, male has better emotion performance than the female. In general, the proposed system has higher performance than the existing system.

Actions (login required)

View Item