Emotion Detection of Thai Elderly Facial Expressions using Hybrid Object Detection

An elderly population is a special group that needs to be taken care of closely. A key area of concern for the elderly is that of mental health and many technologies can be applied in this area. One possible tool is facial expression recognition (FER) that can be used to detect emotions of the elderly for the purpose of mental health care. In this research, we propose a hybrid of Faster R-CNN, SSD, and YOLOv5 object detection models for elderly facial expression detection. In our experiments, the proposed hybrid model was trained on a Thai elderly facial emotion dataset, and its performance was compared to a single-model of Faster R-CNN, SSD, and YOLOv5. The experimental results indicated that the proposed hybrid object detection model had achieved the best performance with an accuracy of 94.07%. This was comparatively better than YOLOv5, which gave the accuracy of 93.33%.