Activity Classifier: A Novel Approach Using Naive Bayes Classification

Abstract

Activity movements have been recognized in various applications for elderly needs, athletes activities measurements and various fields of real time environments. In this paper, a novel idea has been proposed for the classification of some of the day to day activities like walking, running, fall forward, fall backward etc. All the movements are captured using a Light Blue Bean device incorporated with a Bluetooth module and a tri-axial acceleration sensor. The acceleration sensor continuously reads the activities of a person and the Arduino is designed to continuously read the values of the sensor that works in collaboration with a mobile phone or computer. For the effective classification of a person’s activity correctly, Naïve Bayes Classifier is used. The entire Arduino along with acceleration sensor can be easily attached to the foot of a person right at the beginning of the user starts performing any activity. For the evaluation purpose, mainly four protocols are considered like walking, running, falling in the forward direction and falling in the backward direction. Initially five healthy adults were taken for the sample test. The results obtained are consistent in the various test cases and the device showed an overall accuracy of 90.67%.