Studies on Human Recognition Activities Based on Federated Learning
Shuzhen Xu, Yanhong Liu, Xin He · 2022 International Conference on Computer Engineering and Artificial Intelligence (ICCEAI) · 2022
The traditional Human activity recognition (Human Activity Recognition, HAR) failed to protect user data effectively and had low recognition accuracy in the process of training model by machine learning. Under protecting the privacy of user data, improving the accuracy of HAR recognition has become a hot spot. This paper designs the HFL framework for solving human recognition activity based on federated learning and deep learning, and proposes the P-FedAvg algorithm. The P-FedAvg algorithm solves the model bias caused by direct averaging and improves the accuracy of aggregation. Experiments show that our method improves the accuracy of HAR recognition while protecting user's privacy.