Rotation-Invariant Gait Identification with Quaternion Convolutional Neural Networks (Student Abstract)

Bowen Jing, Vinay Uday Prabhu, Angela Gu, John Whaley

AAAI 2021 pp. 15805-15806

doi:10.1609/AAAI.V35I18.17899 /aaai/2021/jing2021aaai-rotation/

Abstract

Accelerometric gait identification systems should ideally be robust to changes in device orientation from the enrollment phase to the deployment phase. However, traditional Convolutional Neural Networks (CNNs) used in these systems compensate poorly for such distributional shifts. In this paper, we target this problem by introducing an SO(3)-equivariant quaternion convolutional kernel inside the CNN. Our architecture (Quaternion CNN) significantly outperforms a traditional CNN in a multi-user gait classification setting. Additionally, the kernels learned by QCNN can be visualized as basis-independent trajectory fragments in Euclidean space, a novel mode of feature visualization and extraction.

PDF AAAI Semantic Scholar

Cite

Text

Jing et al. "Rotation-Invariant Gait Identification with Quaternion Convolutional Neural Networks (Student Abstract)." AAAI Conference on Artificial Intelligence, 2021. doi:10.1609/AAAI.V35I18.17899

Markdown

[Jing et al. "Rotation-Invariant Gait Identification with Quaternion Convolutional Neural Networks (Student Abstract)." AAAI Conference on Artificial Intelligence, 2021.](https://mlanthology.org/aaai/2021/jing2021aaai-rotation/) doi:10.1609/AAAI.V35I18.17899

BibTeX

@inproceedings{jing2021aaai-rotation,
  title     = {{Rotation-Invariant Gait Identification with Quaternion Convolutional Neural Networks (Student Abstract)}},
  author    = {Jing, Bowen and Prabhu, Vinay Uday and Gu, Angela and Whaley, John},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year      = {2021},
  pages     = {15805-15806},
  doi       = {10.1609/AAAI.V35I18.17899},
  url       = {https://mlanthology.org/aaai/2021/jing2021aaai-rotation/}
}