Speech Emotion recognition feature Extraction and Classification

Abstract

Emotion recognition from speech is one of the most popular topics in human computer interaction (HCI). Emotions can be identified by facial expression analysis or by verbal expressions. Several researchers build systems to understand various emotions in human expression. Human emotions automatically recognized by machines for improving human machine interaction. In this article, we have identified three basics of speech emotion recognition system 1) databases 2) feature extraction and 3) various methods of classification. Discussed about the performance of speech emotion recognition system. Features are classified as Essential, Prosodic and Spectral characteristics. Different classifying techniques are used to classify different emotions from human speech like Hidden Markov Model (HMM), K-Nearest Neighbour (KNN),Gaussian Mixtures Model (GMM),Support Vector Machine (SVM) and deep learning.