A Systematic Literature Review on Temporal Graph Based Human Activity Recognition Using Deep learning technique

Human Activity Recognition (HAR) is gaining importance in many real-world applications due to the fast development of AI and deep learning. It is a vital research area aimed at classifying human actions using diverse data modalities, such as images and videos, each contributing uniquely to understanding human behavior. A wide variety of fields can benefit from these, including medicine, smart home security, and user experience improvement. The model for efficient robust HAR in video sequences is designed with Gradient-based Joint Histogram Equalization (GBJHE) thereby improving feature visibility crucial for accurate recognition, a deep convolutional neural network (CNN), is used for feature extraction, capturing detailed hierarchical features from input data. The study includes various datasets used, and high-level deep learning models built thus far and current difficulties suggest future directions for constructing robust and scalable HAR systems for real-world applications.