A new activity classification method K-SVM using Smartphone data

The aim of the Human activity recognition (HAR) is to understand and predict the physical behavior of the human body whether in movement or in a stagnated position. In this paper, we try to improve the classification rate of common physical activities. We design a new classification method K-SVM using the combination between the Weighted K Nearest Neighbour (WKNN) and the Support Vector Machines (SVM). On the other part, the exploitation of the different positions of the sensors on the body is a very interesting research track. We studied the impact of each body part sensor position on the human activity recognition accuracy rate, and then tried to improve the classification rate by using the K-SVM method.