Solving In-door Human Activity Recognition via RFID based on Unsupervised Domain Adaptation

With the changes of the times and the rise of technology, artificial intelligence and radio frequency identification technology (RFID) have become an important part of indoor fields such as health care, smart home, etc., which makes indoor human activity recognition based on RFID and artificial intelligence technology a hot research topic. Due to the complexity and diversity of indoor scenes, there will be large deviations in the recognition of behaviors and actions in different indoor scenes. Based on the above considerations, this paper combines the idea of unsupervised domain adaptation, and uses technologies such as KNN, SVM and FCN to establish a human activity recognition model, thereby improving the accuracy of activity recognition. After applying unsupervised domain adaptation, the performance of all models is greatly improved, among which, the accuracy of the SVM model reaches 90.8% and the F1score reaches 90.4%. With the help of this model, user behavior can be quickly and accurately identified in indoor scenarios such as nursing homes and rehabilitation wards, providing effective security for the daily activities of vulnerable groups such as the elderly and patients, and has broad industry application prospects.