Human Daily Activity Recognition Based On Online Sequential Extreme Learning Machine

Wireless-sensor-network-based health care for human activities involves functional assessment of daily activities. Traditionally, the recognition algorithms adopt batching learning to train network. However, the amount of sensor data is considerable and not all training data arrives together, the learning procedure is time-consuming and the network weights can not be updated online. In this paper, a classifier based on Online Sequential Extreme Learning Machine (OS-ELM) is presented, and used to recognize falling down, running, upstairs, lying down, downstairs, walking, standing and sitting. The system for monitoring human daily activities is designed through a triaxial accelerometer and two pressure sensors in the laboratory and the experiment results are encouraging for human daily activity recognition.