Recently human activity recognition has encouraged a great deal of interest due to its impact on various areas of application. As human’s brain own ability to recognize the actions relies upon obtaining information from a number of senses, not only our vision system; we propose to enhance vision-based activity recognition systems by integrating additional contextual temporal data being sensed from ubiquitous sensors as well. In this paper, we focus upon exploiting time series to extract a user’s daily life patterns and identify any deviations in the temporal patterns. First, we apply Information Gain Temporal Segmentation (IGTS) method, a generic and robust temporal segmentation approach for heterogeneous time series data. From the output of temporal segmentation, daily working patterns are extracted for which unusual user behavior is identified automatically based on high deviation scores of days from the reference transition times. Finally, we evaluate the accuracy of identified days that have such behavior deviations by comparing them with the visual dataset as the ground truth.
@inproceedings{deldari2019inferring,
title={Inferring Work Routines and Behavior Deviations with Life-logging Sensor Data},
author={Deldari, Shohreh and Liono, Jonathan and Salim, Flora D and Smith, Daniel V},
booktitle={Proceedings of ACM International Conference on Web Search and Data Mining (WSDM) workshop on Task Intelligence (TI@ WSDM)(2019). ACM},
year={2019}
}
© 2021 Flora Salim - CRUISE Research Group.