Abstract - The speech signal carries some sort of information, which consists of the intent to be conveyed, who speaks the information, and the emotional information that shows the emotional state of the utterance. One of the characteristics of human voice is the fundamental frequency. In this study the selection of features and methods of classification and recognition is important to recognize the emotional level (anger, sadness, fear, pleasure and neutral) contained in the dataset, this research proposes design through two main processes of training and introduction recognition). Experiments conducted using the Indonesian emotion voice dataset and the Mel-Frequency Cepstrum Coefficients (MFCC) algorithm were used to extract features from sound emotion. MFCC produces 13 cepstral coefficients of each of the sound emotion signals. This coefficient is used as an input of classification of emotional data from 250 data sampling.
Copyrights © 2017