Robust Features for Emotion Recognition from Speech by Using Gaussian Mixture Model Classification

Identification of emotions from speech is a system which recognizes the particular emotion automatically without basing on any particular text or a particular speaker. An essential step in emotion recognition from speech is to select significant features which carry large emotional information about the speech signal, speech signal has an important features. The features extracted from the shape of speech signal are used such as MFCC, spectral centroid, spectral skewness, spectral pitch chroma. These features have been modeled by Gaussian mixture model and optimal number of Gaussians is identified. IITKGP-Simulated Emotion Speech corpus is used as database and four basic emotions such as anger, fear, neutral and happy are considered. The different mixture of spectral features is extracted and experiments were conducted.