Domain Adaptation for Human Fall Detection Using WiFi Channel State Information

Hirokazu Narui, Rui Shu, Félix F. González-Navarro, Stefano ErmonPrecision Health and Medicine 2020 · OpenReview

Abstract

We develop a novel deep learning technique for human fall detection using WiFi Channel State Information (CSI) of a WiFi transmitter and receiver. Different motions in the environment generate distinct features in CSI, which can be fed to a supervised learning machine learning algorithm for training. However, the CSI varies from one environment to another, requiring the collection of environment-specific training data. To overcome this challenge, we propose 1-d convolutional neural network using domain adaptation technique. By adapting to un-labeled data from a new environment, we significantly improve precision and recall, making activity recognition accurate in new environments.
BibTeX
@inproceedings{narui2020domain,
  author = {Hirokazu Narui and Rui Shu and Félix F. González-Navarro and Stefano Ermon},
  title = {{Domain Adaptation for Human Fall Detection Using WiFi Channel State Information}},
  booktitle = {Precision Health and Medicine},
  year = {2020},
  url = {https://openreview.net/forum?id=w2nMvaI8s3},
}

Loading related content...

Figure