Related Experiment Video
Updated: Oct 7, 2025
06:37
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
4.3K
WiPg: Contactless Action Recognition Using Ambient Wi-Fi Signals
Zhanjun Hao1,2, Juan Niu1, Xiaochao Dang1,2
1College of Computer Science and Engineering, Northwest Normal University, Lanzhou 730070, China.
Sensors (Basel, Switzerland)
|January 11, 2022
Summary
This study introduces WiPg, a novel motion recognition model using Wi-Fi channel state information (CSI). WiPg achieves high accuracy in recognizing actions across different individuals, overcoming previous limitations in personnel-dependent systems.
Area of Science:
- Computer Science
- Machine Learning
- Signal Processing
Background:
- Motion recognition has diverse applications, with recent interest in using Wi-Fi channel state information (CSI).
- CSI data during human activity contains rich personal information, limiting the generalizability of motion recognition models across individuals.
- Existing models often fail to perform well when predicting motions of individuals different from those in the training data.
Purpose of the Study:
- To develop a personnel-independent action-recognition model for enhanced motion recognition.
- To address the challenge of model performance degradation due to variations in individual body types and characteristics.
- To achieve accurate motion recognition using Wi-Fi CSI that is robust across different users.
Main Methods:
- Proposed a novel model named WiPg, integrating Convolutional Neural Networks (CNN) and Generative Adversarial Networks (GAN).
- Utilized CSI data collected from 10 experimenters performing 14 distinct yoga movements.
- Trained and tested the WiPg model on this diverse dataset to evaluate its cross-personnel recognition capabilities.
Main Results:
- The WiPg model demonstrated excellent recognition performance, achieving an average correct rate of 92.7% for 14 yoga poses.
- The model successfully achieved 'cross-personnel' movement recognition, indicating independence from the specific individual's body type.
- Experimental results validated the model's effectiveness in generalizing across different users without prior specific training for each.
Conclusions:
- The developed WiPg model offers a robust solution for personnel-independent motion recognition using Wi-Fi CSI.
- WiPg significantly advances the field by enabling accurate action recognition across diverse individuals, overcoming previous limitations.
- This approach holds promise for various applications requiring reliable and generalized human motion analysis via wireless signals.
