Main Article Content
Currently, the world’s elderly have increased substantially. The elderly living alone are prone to accidents such as falls, which sometimes lead to fatalities without timely notifications and help. This research applies data mining techniques in order to classify activities done by the elderly, including lying, sitting, standing, walking, running, and fall, based on the data obtained from a smartphone’s sensors. We collected data from a sample of 25 elderly volunteers and separated the data into three data sets. Values of the first data set were from the three accelerometer axes, X, Y and Z as well as the Euclidean distance between the data samples. The values of the second data set were the angle values, X, Y and Z obtained from the sensors’ fusion of accelerometer, magnetometer and gyroscope as well as the distance between the data samples. The third data set, however, used the first data set, second data set, and the distances between the data samples. Thereafter, the three data sets were used to find the best classifier for our application, from neural network (Multi-layer Perceptron), Naive Bayes, Support Vector Machine, and K-Nearest Neighbor. The application was able to track the elderly’s activities in real-time and send notifications when falling or having an undue activity time. The neural network (multi-layer perceptron) had the highest accuracy for the activity classifications of 99.52%.
Daily Life Activities, Sensor Fusion, Activity Classifications, Elderly
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.