Motion Recognition from Accelerometer, Gyroscope and ECG Data



In the last years, with the advances in smart personal things, make us to believe that the new generation of Internet of Things will become the human being as an integral part of the system. With this purpose the IoT technologies must be supported on Human-in-the-Loop Systems. In fact, the Human-in-the-Loop Cyber-Physical System (HiLCPS) considers human being as an integral part of the system. The data acquisition and sensing are fundamental parts within HiLCPS. In the data acquisition process, the principal devices that allow data collection are the sensors. There are several sensors that can be used to gather data on human activity and behavior such as the microphone, accelerometer, gyroscope, magnetometer, ECG, etc. With data collected is possible to infer various activities and moods of individuals and based on this information it is also possible to generate a feedback to improve quality life of a person. The reliability of data acquired is crucial, for this reason, the selection of a sensor or smart device is very important, and it must be taken depending on the type of applications to be implemented. Combining information from various sensors can represent an opportunity to improve the reliability and the accuracy of data. In this paper, we analyzed the accuracy of various classifiers for motion recognition. A public dataset that includes data from various sensors was analyzed. The data of accelerometer, gyroscope and electrocardiogram (ECG) were used in order to interfere human activity. The performance was evaluated applying the following classifiers: Decision Trees, Support Vector Machine (SVM), K-Nearest Neighbors (K-NN) and Hybrid Classifier. Finally, for the recognition of the four movement activities (no, slight, moderate and high movement), we obtained an accuracy of 99,40% with the K-NN classifier.


22nd Conference RecPad 2016, Aveiro 2016

Cited by

No citations found