Authors: Cunyi Yin, Xiren Miao, Jing Chen, Hao Jiang, Deying Chen, Yixuan Tong, Shaocong Zheng

Published on: March 05, 2024

Impact Score: 7.8

Arxiv code: Arxiv:2403.02632

Summary

• What is new: A novel semi-supervised cross-domain neural network (SCDNN) for low-resolution infrared-based human activity recognition (HAR) is introduced, which outperforms existing deep learning methods in cross-domain scenarios.
• Why this is important: Low-cost and private human activity recognition in varying environments poses limitations with existing technology.
• What the research proposes: SCDNN uses a feature extractor, domain discriminator, and label classifier for better domain adaptation and activity recognition with minimal labeling effort.
• Results: The method achieves a 92.12% accuracy in recognizing activities in new environments, demonstrating its adaptability and low-cost effectiveness.

Technical Details

Technological frameworks used: SCDNN based on low-resolution infrared sensor

Models used: Feature extractor, domain discriminator, label classifier

Data used: Unlabeled and minimal labeled target domain data

Potential Impact

Low-cost surveillance, smart home devices manufacturers, healthcare monitoring companies.

Want to implement this idea in a business?

We have generated a startup concept here: InfiMotion.