Based Human Activity Recognition

Wearable sensor-based human activity recognition (HAR) aims to automatically identify human actions using data from devices like smartwatches and fitness trackers. Current research emphasizes improving efficiency and accuracy through techniques like knowledge distillation to reduce model complexity, multimodal sensor fusion (e.g., combining inertial measurement units and ambient light sensors), and the application of advanced architectures such as transformers and recurrent neural networks. These advancements are crucial for developing more robust and energy-efficient HAR systems with applications in healthcare, assisted living, and industrial settings, particularly addressing challenges like data scarcity and noisy labels in real-world scenarios.