Optimal Methods for Activity Recognition using Machine Learning Techniques

  • B. Mounika
  • B. Mamatha
  • P. Anjaiah


Action recognition is a method of monitoring the activities performed by the individual so that it will predict the disorders prior that causes in the future. This can be monitored to the human at various locations by body-worn devices. Worn devices will record the data of the physical activity based upon the movements created by the subjects. We have collected the dataset from UCI Machine learning repository. The dataset collected by conducting the experiment on nearly thirty subjects by performing six activities like walking, walking upstairs, walking downstairs, sitting, standing and lying with a tagged smartphone on a waist location. The associated sensors accelerometer and gyroscope sensor dataset extracted from the smartphone and applied machine learning methods to recognize the physical activity and to evaluate the performance variations. The accuracy obtained by the machine learning techniques is high compared to accelerometer. In this research, we compare physical activity with accelerometer and machine learning. For example, we can consider some physical activities like walking upstairs and walking downstairs, etc., activities force and stress levelsincreases, we found the levels with Machine Learning methods.