A CNN-Based Approach for Facial Expression Recognition in Mentally Retarded Individuals

This paper proposes a facial expression recognition (FER) system of a Convolutional Neural Network (CNN) model which recognizes seven emotions which are Anger, Disgust, Fear, Happiness, Sadness, Surprise, and Neutral for mentally retarded persons.The figure below demonstrates the proposed model which has been trained and tested on FER2013 giving us a validation accuracy of ~93%.It used robust feature extraction with convolutional and pooling layers, which allowed for a high accuracy of clear expressions and an improved handling of subtle emotions.Dropout layers were used to reduce overfitting.These results suggest the feasibility of employing the FER system in applications like behavioral tracking, therapeutic assistive, and rehabilitation applications.Further, our future work will primarily focus on transfer learning, multimodal emotion recognition, and real-time deployment.