Emotion recognition of the SROL Romanian database using fuzzy KNN algorithm

This study is focus on the supervised algorithm in order to classify the emotions from speech. The fuzzy-KNN classifier algorithm comparing with the classical KNN has the advantage to quantify the “strength” of the membership to a class. In the classical KNN algorithm, the decision regarding the assigning of an instance to a class was taken only based on the majority number of neighbors in a particular class; each neighbor has the same importance in the classification process. Therefore the results obtained with fuzzy KNN algorithm are improved compared to those obtained in our previous studies. This paper aims to analyze the percentages of the emotion classification using statistical parameters extracted from the SROL emotional database. The features vectors contain 17 parameters; in the future we intend to extend the number of parameters used for classification of the emotions.