Publications
カンファレンス (国際) The Power of Periodicity: Exploiting Periodic UWB CIRs for Robust Activity Recognition with Attention-aware Multi-level Wavelet
Han Lin (Tokyo Institute of Technology), Atsushi Nomura (Tokyo Institute of Technology), Kota Tsubouchi, Nobuhiko Nishio (Ritsumeikan University), Masamichi Shimosaka (Tokyo Institute of Technology)
The 23rd International Conference on Pervasive Computing and Communications (PerCom 2025)
2025.4.1
As a promising device-free solution, wireless sensing techniques, such as leveraging Wi-Fi or UWB signals, have been applied to various applications, addressing privacy concerns associated with traditional computer vision-based methods and freeing people from the inconvenience of carrying devices. Limited research has explored the use of UWB CIR for activity recognition, which shows improved performance over the popular Wi-Fi CSI. However, prior studies overlooked the brevity of the single measurement and have not fully exploited the potential of feature extraction from CIRs. This research propose a robust device-free activity recognition approach using UWB CIRs, which employs multi-level WPD and customed attention layer to selectively combine multi-resolution features, , thereby reducing the cost of fine tuning and enhancing recognition accuracy. Experiments conducted in several typical scenarios validate the performance of the proposed method. The ablation study with various time-frequency resolutions and comparison with STFT highlight the necessity of MRA in device-free HAR and the cost-reduction of our proposal.
Paper :
The Power of Periodicity: Exploiting Periodic UWB CIRs for Robust Activity Recognition with Attention-aware Multi-level Wavelet
（外部サイト）