Abstract

In many applications like emotional analysis, human- computer interaction, mental health monitoring, sentiment analysis and surveillance systems, real-time facial expression detection has become a vital role. Real-time facial expression recognition systems recognize human emotions which improves user experiences and system reactions. The Convolutional Neural Network (CNN) algorithm with three convolution layers is used for human expression recognition. Two different datasets, FER2013 and CK +48 are used for training the proposed system. This dataset provides a diverse range of facial expressions for training and evaluation. The proposed system has trained for seven distinct emotions: anger, sadness, happiness, neutral, disgust, surprise, and fear. Many existing systems are accurate but suffer from complexity in their model architecture and code model implementation. The proposed system achieves a notable accuracy of 92%, outperforming many existing models in the field. The proposed model has high accuracy with less complexity which is suitable for real time deployment. The proposed solution streamlines the design while preserving performance, resulting in greater ease of utilization and less computation requirements.

Authors

Purva Kalambate, Seema Hanchate, Poonam More, Janhavi Askar
SNDT Women’s University, India

Keywords

Facial Expression, CNN, Feature Extraction, Emotions Facial Landmarks, Emotion Detection