Sensor Based Human Activity Recognition

Sensor-based human activity recognition (HAR) aims to automatically identify human actions using data from wearable sensors, primarily to improve healthcare, sports performance monitoring, and human-computer interaction. Current research emphasizes improving model accuracy and robustness by addressing data heterogeneity, utilizing advanced architectures like convolutional neural networks (CNNs), recurrent neural networks (RNNs, especially LSTMs), and transformers, often incorporating techniques such as multi-modal fusion, self-supervised learning, and transfer learning to overcome limitations of labeled data scarcity. These advancements hold significant potential for creating more personalized and efficient systems for various applications, particularly in areas requiring continuous, unobtrusive monitoring of human behavior.