Speech Based Emotion Recognition

This paper presents a method to identify the emotion of an audio segment with an intention to recognize human emotional/mental status. Four features namely energy, pitch, Formants, Mel frequency cepstral coefficients (MFCC) and their derivatives are used to recognize emotions such as fear, anger, happiness and sadness. PCA is used to reduce the feature dimensionality. Support vector machine is implemented to perform the emotional state classification. The overall recognition rate obtained is 84.99% using samples of Berlin emotional speech database. Keywords—MFCC, Formants, Pitch, Energy, PCA, Support Vector Machine (SVM)