Invasive Device Human Activity Recognition based on Smartphone Dataset
TL;DR
This study used accelerometer and gyroscope data to train the proposed SVM and Random Forest for human activity recognition and confirmed the accuracy, F1 score, precision, recall and visualize confusion matrix of the proposed HAR model.
Abstract
The Human Activity Recognition (HAR) methodology serves essential functions within healthcare settings as well as sports disciplines and applications in both human-computer communication and smart system environments. This paper presents invasive human activity recognition based on smart phone datasets. In this study, we used accelerometer and gyroscope data to train the proposed SVM and Random Forest for human activity recognition. The Human Activity Recognition with smartphone public dataset is utilized to estimate the accuracy, F1 score, precision, recall and visualize confusion matrix of the proposed HAR model. The Human Activity Recognition database was built from the recordings of 30 study participants performing activities of daily living (ADL) while carrying a waist-mounted smartphone with embedded inertial sensors. The experiments have been carried out with a group of 30 volunteers within an age range of 19-48 years. Each person performed six activities (walking, walking upstairs, walking downstairs, sitting, standing, laying) wearing a smartphone (Samsung galaxy s ii) on the waist. Training SVM with RBF kernel confusion matrix with Accuracy: 0.9884, F1 Score: 0.9884, Recall: 0.9884 and Precision: 0.9885Training Random Forest, confusion matrix with Accuracy: 0.9742, F1 Score: 0.9741, Recall: 0.9742 and Precision: 0.9745 was achieved.