Abstract –In this study, we develop a FER system for intellectual disability people based on DWT and Gabor filter for feature extraction followed by an SVM classifier for classification of expression. Four different preprocessing techniques such as converting into grayscale, normalizing the image, and data augmentation were performed on the MuDERI dataset to improve the accuracy. However, the accuracy of the proposed model is 83.87% showing its ability for recognizing unique facial expression in it. This will enable and explore integration with deep learning and optimization strategies, real-time implementation, and adaptable.
