Integrated Sensory Perception: Real-Time Object Detection, Action Recognition and Emotion Analysis

- This research study presents a comprehensive approach to emotion recognition leveraging a combination of object detection, action recognition, and emotion recognition models, integrated with Kinect sensors for accurate mapping in low-light environments. The model achieves a combined precision and recall range of 82-85%, demonstrating its effectiveness in identifying objects and emotions. Object detection is performed using YOLO, action recognition utilizing LSTM, and emotion recognition employing CNN. Additionally, low-light robustness is ensured through the incorporation of a Zero-DCE model. A unique dataset for training the model under low-light conditions is generated using Kinect sensors. The methodology involves rigorous analysis of behavior, encompassing object detection and tracking methods, gesture and emotion recognition, data preprocessing, and feature extraction. Findings indicate that YOLO-based tracking approaches significantly enhance emotion recognition systems. This study contributes to advancing behavior analysis by highlighting the efficacy of various object tracking techniques. particularly in low-light environments.