A CLAHE-Enhanced CNN Approach to Facial Emotion Recognition as Assistive Technology for the Visually Impaired

Visual limitations in visually impaired individuals pose challenges in understanding facial expressions that are important in social interactions. This research develops a CNN-based face detection system equipped with CLAHE preprocessing to improve image quality, especially in environmental conditions with diverse lighting. The system is designed to support the visually impaired in recognizing facial expressions through seven emotion classes. The model was trained using a combined FER2013 dataset and a self-built dataset with Indonesian faces. Evaluation results show that the model is able to recognize positive expressions quite well, although it still has difficulties distinguishing between similar negative emotions. With CLAHE preprocessing, the model achieved an accuracy of 61.51%, indicating improved performance under varied lighting. This system has the potential to assist the visually impaired in improving social interaction through automatic facial expression recognition in various lighting conditions.