Activity Classification

Activity classification aims to automatically identify human actions from various data sources, such as wearable sensors, video, audio, and radar, primarily to improve healthcare monitoring, understand human behavior, and enhance human-computer interaction. Current research emphasizes developing robust and data-efficient models, exploring both supervised and self-supervised learning approaches with architectures including convolutional neural networks, recurrent neural networks (like LSTMs and GRUs), and transformer-based models, often incorporating techniques like domain adaptation and sensor fusion. This field is significant for its potential to improve personalized healthcare, create more intuitive assistive technologies, and advance our understanding of human movement and behavior across diverse contexts.