Human Activity Recognition using WISDM Datasets

Abstract

In this research, we describe a method for identifying human activity where we use a model that comprises of Convolutional Neural Network (CNN), using the datasets acquired from smartphones. Jogging, sitting, walking, standing, going downstairs and upstairs are included in the daily activities that taken into consideration. We use the WISDM datasets acquired from the three-dimensional raw accelerometer sensors. The performance of our CNN model showed 87.85% accuracy. This performed better than the Support Vector Machine model, which had an accuracy rate of 82.27 percent. Therefore, the results demonstrate that our suggested strategy can outperform current best practices for human activity recognition.