8th International Conference on Body Area Networks

Research Article

Monitor Pilgrims: Prayer Activity Recognition using Wearable Sensors

  • @INPROCEEDINGS{10.4108/icst.bodynets.2013.253685,
        author={Amir Muaremi and Julia Seiter and Franz Gravenhorst and Agon Bexheti and Bert Arnrich and Gerhard Troester},
        title={Monitor Pilgrims: Prayer Activity Recognition using Wearable Sensors},
        proceedings={8th International Conference on Body Area Networks},
        keywords={wearable sensors pilgrims prayer hrv gsr respiratory},
  • Amir Muaremi
    Julia Seiter
    Franz Gravenhorst
    Agon Bexheti
    Bert Arnrich
    Gerhard Troester
    Year: 2013
    Monitor Pilgrims: Prayer Activity Recognition using Wearable Sensors
    DOI: 10.4108/icst.bodynets.2013.253685
Amir Muaremi1,*, Julia Seiter1, Franz Gravenhorst1, Agon Bexheti2, Bert Arnrich3, Gerhard Troester1
  • 1: Wearable Computing Lab, ETH Zurich
  • 2: Artificial Intelligence Lab, EPFL
  • 3: Computer Engineering Department, Bogazici University
*Contact email: muaremi@ife.ee.ethz.ch


Each year, millions of Muslims visit the holy sites in Makkah and Madinah. The so-called Hajj pilgrimage is one of the biggest annual events in the world. The individual impact on participating pilgrims is significant, with many of the pilgrims reporting it as a life-changing event. However, quite a little is done to objectively monitor the pilgrims and to understand, from the individual point of view, the characteristics of each of the pilgrimage stages. In this work, through observing differences in bio-physiological responses of the subjects during prayers, we are able to differentiate, in the first case, between congregational prayers and individual prayers, and, in the second case, between silent prayers and loud prayers. We collected data from 10 participants in an 8-day pilgrimage using two wearable sensors, namely chest belts and wrist-worn devices. We derive features from ECG, respiration and GSR data, and use the ANOVA model to analyze feature groups. Based on that, we build classifiers to differentiate between types of prayers. The SVM classifier shows the best performance with a mean accuracy rate of 78 % for the first case, and 84 % for the second case.