2026-03-14T23:03:04Z https://ipsj.ixsq.nii.ac.jp/oai

oai:ipsj.ixsq.nii.ac.jp:00194582 2025-01-19T23:25:48Z 1164:3865:9692:9714

様々な人体動作に対するパイスタティック散乱特性の大規模データセットの合成 Towards motion recognition by ubiquitous RF sensing: a large synthetic dataset of bistatic scattering from human body in motion 龐, 海涛 高田, 潤一 Haitao, Pang Jun-ichi, Takada This paper discusses using electromagnetic (EM) signals of wireless communication for human motion recognition (HMR) via radio frequency (RF) sensing. Modern high data-rate RF communication systems have inherent sensing capabilities for moving objects. Contact-less, device-less, camera-less HMR by ordinary radio signals is very advantageous compared with traditional method by computer vision or wearables. However, the physical phenomena of RF scattering from biological bodies are complicated and dependent on many factors such as shape, material properties, polarization, geometry, etc. To demonstrate this, we focus on the characteristics of time variant scattering from a human body performing various kinds of motions, in a data-centric manner. Using inputs of measured motion data, generative human models in various postures, and a numerical EM simulator based on a high frequency asymptotic method of physical optics (PO), large datasets can be generated. This can be used for better and more comprehensive understanding of the RF interaction with human body, and for development and comparison of future robust recognition RF algorithms and joint communication and radar (RAdio Detection And Ranging) devices. This paper discusses using electromagnetic (EM) signals of wireless communication for human motion recognition (HMR) via radio frequency (RF) sensing. Modern high data-rate RF communication systems have inherent sensing capabilities for moving objects. Contact-less, device-less, camera-less HMR by ordinary radio signals is very advantageous compared with traditional method by computer vision or wearables. However, the physical phenomena of RF scattering from biological bodies are complicated and dependent on many factors such as shape, material properties, polarization, geometry, etc. To demonstrate this, we focus on the characteristics of time variant scattering from a human body performing various kinds of motions, in a data-centric manner. Using inputs of measured motion data, generative human models in various postures, and a numerical EM simulator based on a high frequency asymptotic method of physical optics (PO), large datasets can be generated. This can be used for better and more comprehensive understanding of the RF interaction with human body, and for development and comparison of future robust recognition RF algorithms and joint communication and radar (RAdio Detection And Ranging) devices. technical report 情報処理学会 2019-02-25 application/pdf 研究報告モバイルコンピューティングとパーベイシブシステム（MBL） 30 2019-MBL-90 1 5 2188-8817 AA11851388 https://ipsj.ixsq.nii.ac.jp/record/194582/files/IPSJ-MBL19090030.pdf eng