LightBGM for Human Activity Recognition Using Wearable Sensors
Zhaosheng Shao, Jianxin Guo, Yushuai Zhang, Rui Zhu, Liping Wang · 2021
Human activity recognition (HAR) has become a hot research topic, especially in the fields of vital records, adaptive tracking and health monitoring. In order to overcome the problem of low accuracy of commonly used algorithms for human recognition activities, this paper proposes a LightBGM classification method based on the UCI dataset. This method combines the user and the surrounding environment with the computer, and uses the smart phone to perceive people’s actions. It does not need to use special sensors to collect the action information in different positions of the body. It only uses the smart phone with inertial sensors to collect the action information, and carries out classification and recognition, and tests on UCI data sets. The LightGBM algorithm shows a higher accuracy rate than other algorithms and can more accurately identify various actions.