Go direct to the page content

Deep HAR

Deep HAR

... Various Human Activities

Deep Human Activity Recognition

Human Activity Recognition (HAR) detects the kind of movement a person has made.

Since omni-directional antenna and multiple paths, human movements change the signals received by the receiver.

  • Analyzes the patterns of changes in wireless signals, extracts unique features and classifies the user's behavior.
Research Topic

Multi-path Separation

When a user changes his/her position and orientation, the multi-paths from Tx to Rx alter, which affects the changing pattern in wireless signals. We are trying to separate multi-path at receiver-side to harvest more information from CSI.

Superimposed Signal for Multipath Separation
Research Topic

Domain Adaptation for Robustness

The multi-paths completely changes at new spaces. It is impossible to collect all the data necessary to account for such variation of surrounding environment. To address the challenge, we are developing DL model which only requires limited samples for domain adaptation.

Unsupervised DA and Few-shot DA at CSI-based HAR