Emotion Detection and Gender Recognition from Speech Signal using Deep Learning

Abstract

Emotions are subjective and it is normal for different people to interpret them in their own ways. It is
a major task to find datasets with audios which are not biased and the main datasets present are from
news channels or movie clips. In our project, we would be using the dataset of RAVDESS which
contains 24 professional actors’ audio sets. It also determines the gender of the actors through
MFCC, also called state-of-the-art, which accurately represents the short time spectrum manifesting
the shape of the vocal tract. Speech sentiments include disgust, calm, angry, happy, surprise, sad,
neutral and fearful. We would be developing a convolutional neural network model for emotion
recognition and gender detection from speech. The model would be trained with the dataset obtained
and would be tested with external audio