Research on Behavior Recognition of Wearable Devices Based on Inertial Sensors

With the increasing demand for intelligent interaction and health monitoring, human behavior recognition based on acceleration sensors has attracted extensive attention in the fields of medical care, motion detection, energy consumption assessment and so on. MPU9250 was used to extract acceleration signals and angular velocity signals in human behavior. The recognition results are obtained through the steps of signal preprocessing, feature selection, feature extraction and recognition algorithm. The behavior recognition method based on decision tree is used to identify the stillness, walking, running and jumping.