Supervised Human Activity Recognition

Supervised Human Activity Recognition (HAR) aims to automatically classify human actions from sensor data, such as wearable sensors or video, primarily focusing on improving accuracy and reducing reliance on large labeled datasets. Current research emphasizes semi-supervised and self-supervised learning techniques, employing deep learning models like Siamese networks and contrastive learning methods to leverage unlabeled data and address data scarcity issues. These advancements are significant for applications ranging from healthcare monitoring and assistive robotics to security and human-computer interaction, offering more robust and efficient activity recognition systems.