EMOTION DETECTION USING VIDEO AUDIO AND TEXT

Abstract

Emotion recognition is essential to improving human-computer interaction in the current era of artificial intelligence. This study introduces a Video and Voice-Based Emotion Detection System that uses textual content, speech tone, and facial expressions to properly identify human emotions. The suggested system combines Long Short-Term Memory (LSTM) networks for voicebased and text-based emotion analysis, Convolutional Neural Networks (CNN) for videobased face emotion identification, and a comparative performance evaluation between these modalities. In order to ensure high accuracy and resilience in a variety of emotional states, including happy, rage, sadness, surprise, and neutrality, the system is built to analyse real-time data from many sources, including text, audio, and video. According to experimental findings, CNN models are more accurate at classifying facial emotions, whilst LSTM networks are better in capturing sequential