Affect Detection from Speech using Deep Convolutional Neural Network Architecture

From last few years there are many research works have been done on the field of affect detection or emotion recognition from speech signal. Researchers has been directed to find out different emotional content from speech signals, they have tried to extract different features from speech and used different types of supervised or unsupervised learning methods to train a network such a way that a model can be developed which can identify emotion from speech signal successfully. The primary challenges of emotion recognition are choosing the emotional speech corpus (speech database), identification of different features related to speech and an appropriate choice of a classification model. In this work we have explored RML emotional speech corpus for our experiment purpose it is a collection of emotional audiovisual files of different languages. We have analyzed the performance of Deep Convolutional Neural Network with Mel-Spectrogram as features for recognition of emotion.