Author : Nagama B. Deshmukh and Chandu R. Barde

The quantity of more established individuals in western nations is always expanding. A large portion of them like to live autonomously and are defenseless to fall occurrences. Falls frequently lead to genuine or even deadly wounds which are the main source of death for elderlies. To address this issue, it is basic to create hearty fall discovery frameworks. In this specific situation, we build up an AI structure for fall location and every day living action acknowledgment. We use speeding up and precise speed information from two open databases to perceive seven unique exercises including falls and exercises of everyday living. From the increasing speed and rakish speed information, we separate time and recurrence area include and give them to an order calculation. In this work, we test the exhibition of four calculations for arranging human exercises. These calculations are fake neural system (ANN), K-closest neighbors (KNN), quadratic help vector machine (SVM), and troupe sacked tree (EBT). New highlights that improve the presentation of the classifier are removed from the force otherworldly thickness of the speeding up. In an initial step, just the quickening information is utilized for movement acknowledgment.

Keywords: Fall detection, activity recognition, machine learning, acceleration data, angular velocity data, feature extraction.