Continual Activity Recognition with Generative Adversarial Networks

Continual learning is an emerging research challenge in human activity recognition (HAR). As increasing number of HAR applications are deployed real-world environments, it important and essential to extend the model adapt change people’s routine. Otherwise, can become obsolete fail deliver activity-aware services. The existing has focused on detecting abnormal sensor events or new activities, however, extending currently under-explored. To directly tackle this challenge, we build recent advance area lifelong machine design a continual system, called HAR-GAN , grow over time. does not require prior knowledge what classes might be store historical data by leveraging use Generative Adversarial Networks (GAN) generate previously learned activities. We have evaluated four third-party, public datasets collected binary sensors accelerometers. Our extensive empirical results demonstrate effectiveness shed insight future challenges.