Neural Networks Based Prayer Monitoring and Recognition Framework

Abstract

Daily prayer is an important part of every muslim around the globe and muslims makes up 24.7% of the world population. The rapid advancement in smartphones and wearable devices sensors', enabled research to utilize these sensors for more complex human activity recognition tasks. In this paper, we propose a framework that leverages smartwatch sensors for prayer monitoring and recognition. Our work is mainly targeting smartwatch due their convenience for the users, but challenging nature of unlimited movement and noise during prayer. Our framework allows user to store their prayer using a smartwatch with the help of our web applications. We further propose a deep learning-based timeseries data classification model for prayer recognition. In this study, we have used three sensors namely heartrate electrocardiogram, an accelerometer, and a gyroscope. The collected data is used to train a prayer classification model. The raw prayer data is pre-processed using moving average, data aggregation, and data segmentation. Our proposed prayer classification model showed an accuracy of 87%. Finally, we discussed a suggested future work and further research in prayer analysis area.