Skip to main content
placeholder image

Human body imaging method based on channel state information and deep learning

Conference Paper


Abstract


  • Human perception technology has been utilized extensively in various fields. WiFi equipment has grown increasingly popular benefited from the rapid growth of wireless communication and the Internet of things, accompanied with an abundance of methods for person recognition based on WiFi channel state information (CSI). This study proposes a method of human body intelligent perception based on DCGAN to address the current issue of image blur generation in CSI-based imaging of the human body. The method mainly comprises the following steps: (i) obtaining CSI data for actual interior scenes, gathering scene video data based on time stamps, and creating heterogeneous data sets from the video extraction mask images collected; (ii) putting heterogeneous data sets into the DCGAN network for iterative training; (iii) obtaining a generation network for regenerating the environment mask image via the CSI data. Experiments demonstrate that such a method requires less space and generates a human body image more quickly, and that the IOU index of the generated image can be increased from 0.66 of the person-in-WiFi method to 0.696, allowing it to generate a clearer human body mask image than the prior network.

Publication Date


  • 2022

Citation


  • Pang, C., Wang, P., Wang, L., Zheng, Q., & Zhu, L. (2022). Human body imaging method based on channel state information and deep learning. In Proceedings of SPIE - The International Society for Optical Engineering Vol. 12503. doi:10.1117/12.2657075

Scopus Eid


  • 2-s2.0-85144211456

Web Of Science Accession Number


Volume


  • 12503

Issue


Place Of Publication


Abstract


  • Human perception technology has been utilized extensively in various fields. WiFi equipment has grown increasingly popular benefited from the rapid growth of wireless communication and the Internet of things, accompanied with an abundance of methods for person recognition based on WiFi channel state information (CSI). This study proposes a method of human body intelligent perception based on DCGAN to address the current issue of image blur generation in CSI-based imaging of the human body. The method mainly comprises the following steps: (i) obtaining CSI data for actual interior scenes, gathering scene video data based on time stamps, and creating heterogeneous data sets from the video extraction mask images collected; (ii) putting heterogeneous data sets into the DCGAN network for iterative training; (iii) obtaining a generation network for regenerating the environment mask image via the CSI data. Experiments demonstrate that such a method requires less space and generates a human body image more quickly, and that the IOU index of the generated image can be increased from 0.66 of the person-in-WiFi method to 0.696, allowing it to generate a clearer human body mask image than the prior network.

Publication Date


  • 2022

Citation


  • Pang, C., Wang, P., Wang, L., Zheng, Q., & Zhu, L. (2022). Human body imaging method based on channel state information and deep learning. In Proceedings of SPIE - The International Society for Optical Engineering Vol. 12503. doi:10.1117/12.2657075

Scopus Eid


  • 2-s2.0-85144211456

Web Of Science Accession Number


Volume


  • 12503

Issue


Place Of Publication