2024-03-28T20:07:16Zhttps://ipsj.ixsq.nii.ac.jp/ej/?action=repository_oaipmhoai:ipsj.ixsq.nii.ac.jp:002071442023-04-27T10:00:04Z01164:03865:10114:10346
Domain-Adversarial Training for Door Event Detection Using Wi-Fi Channel State Informationドメイン敵対学習を用いたWi-Fi CSIによるドアイベント認識enghttp://id.nii.ac.jp/1001/00207042/Technical Reporthttps://ipsj.ixsq.nii.ac.jp/ej/?action=repository_action_common_download&item_id=207144&item_no=1&attribute_id=1&file_no=1Copyright (c) 2020 by the Information Processing Society of Japan現在,大阪大学大学院情報科学研究科現在,NTTコミュニケーション科学基礎研究所現在,大阪大学大学院情報科学研究科現在,大阪大学大学院情報科学研究科現在,NTTアクセスサービスシステム研究所現在,NTTアクセスサービスシステム研究所Kim, Heng尾原, 和也前川, 卓也原, 隆浩村上, 友規アベセカラ, ヒランタDoor event detection has been actively studied as it has many applications such as heating, ventilation, air conditioning control, home automation, and intrusion detection, etc. However, existing method on door event detection using Wi-Fi signals rely on a large amount of training data collected in a target environment. In this paper, we present a deep learning-based method for door event detection using domain-adversarial training to extract environment-independent features of door events from Wi-Fi CSI. It can recognize door events without employing labeled training data collected in a target environment. To achieve recognition across different environments, we leverage domain-independent features of door events, namely, differential and dynamic event features, which capture inherent changes in signal propagation caused by door events regardless to the environment. We evaluated the effectiveness of our proposed method through experiments in real environments. The experimental results demonstrated that the method can achieve the state-of-the-art performance without using labeled training data from a target environment.Door event detection has been actively studied as it has many applications such as heating, ventilation, air conditioning control, home automation, and intrusion detection, etc. However, existing method on door event detection using Wi-Fi signals rely on a large amount of training data collected in a target environment. In this paper, we present a deep learning-based method for door event detection using domain-adversarial training to extract environment-independent features of door events from Wi-Fi CSI. It can recognize door events without employing labeled training data collected in a target environment. To achieve recognition across different environments, we leverage domain-independent features of door events, namely, differential and dynamic event features, which capture inherent changes in signal propagation caused by door events regardless to the environment. We evaluated the effectiveness of our proposed method through experiments in real environments. The experimental results demonstrated that the method can achieve the state-of-the-art performance without using labeled training data from a target environment.AA11851388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)2020-MBL-9633192020-09-222188-88172020-10-01