Human movement recognition system using R

Abstract

With the proliferation of ubiquitous computing, the desire to make everyday life smarter and easier is growing more and more. Human Activity Recognition (HAR) is the result of a similar motivation. By recognizing user activity, HAR enables a wide range of comprehensive IT applications. To contribute to the multifarious applications proposed by HAR, it is essential to plan the appropriate activities. The simplest of the problems is using the wrong data manipulation and the execution of the prediction using erroneous algorithms interfering with the performance of the HAR system. R has proven to be a powerful and flexible tool for data mining and analysis. Here, we analyze the set of data extracted from UCI (University of California, Irvine) dataset using R. As a result of analysis any activity performed by participants will be recognized. As a sample, we are extracting data of 30 volunteers aged 19 to 48, each carrying a Smartphone at the waist. They perform various activities and record the data. Using the confusion matrix to apply on Support Vector Machine algorithm, we extract energy needed to perform activities, the frequency of each domain, etc., from dataset and display the results; standing, lying, or sitting. In other words, we classify the activities to be done by participants. Its applications include surveillance systems, patient monitoring systems and various systems, including interaction between people and electronic devices. This document will drive future research in more productive areas.