The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
Exibições de texto completo
94
A chegada da era da Internet das Coisas (IoT) garantiu a onipresença das tecnologias de detecção humana. As câmeras tornaram-se instrumentos baratos para detecção humana e têm sido cada vez mais utilizadas para esse fim. Como as câmeras produzem grandes quantidades de informações, elas são ferramentas poderosas de detecção; no entanto, como as imagens das câmaras contêm informações que permitem a identificação pessoal dos indivíduos, a sua utilização apresenta riscos de violações da privacidade pessoal. Além disso, como os eletrodomésticos prontos para IoT estão conectados à Internet, imagens capturadas por câmeras de usuários individuais podem vazar involuntariamente. Ao desenvolver nosso método de detecção humana [33], [34], propusemos técnicas para detectar humanos a partir de imagens pouco claras nas quais os indivíduos não podem ser identificados; no entanto, uma desvantagem deste método era a sua incapacidade de detectar seres humanos em movimento. Assim, para permitir o rastreamento de humanos mesmo que as imagens estejam desfocadas para proteger a privacidade, introduzimos uma estrutura de filtro de partículas e propomos um método de rastreamento humano baseado na detecção de movimento e na detecção de frequência cardíaca. Também mostramos como o uso de imagens integrais [32] pode acelerar a execução de nossos algoritmos. Em testes de desempenho envolvendo imagens pouco nítidas, o método proposto produz resultados superiores aos obtidos com o método de desvio de média existente ou com um método de detecção de face baseado em características do tipo Haar. Confirmamos a aceleração proporcionada pelo uso de imagens integrais e mostramos que a velocidade do nosso método é suficiente para permitir a operação em tempo real. Além disso, demonstramos que o método proposto permite um rastreamento bem-sucedido mesmo em casos em que a postura do indivíduo muda, como quando a pessoa se deita, situação que surge em ambientes de uso do mundo real. Discutimos as razões por trás do comportamento superior do nosso método em testes de desempenho em comparação com outros métodos.
Toshihiro KITAJIMA
Samsung R&D Institute Japan
Edwardo Arata Y. MURAKAMI
Samsung R&D Institute Japan
Shunsuke YOSHIMOTO
Osaka University
Yoshihiro KURODA
Osaka University
Osamu OSHIRO
Osaka University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copiar
Toshihiro KITAJIMA, Edwardo Arata Y. MURAKAMI, Shunsuke YOSHIMOTO, Yoshihiro KURODA, Osamu OSHIRO, "Privacy-Aware Human-Detection and Tracking System Using Biological Signals" in IEICE TRANSACTIONS on Communications,
vol. E102-B, no. 4, pp. 708-721, April 2019, doi: 10.1587/transcom.2018SEP0006.
Abstract: The arrival of the era of the Internet of Things (IoT) has ensured the ubiquity of human-sensing technologies. Cameras have become inexpensive instruments for human sensing and have been increasingly used for this purpose. Because cameras produce large quantities of information, they are powerful tools for sensing; however, because camera images contain information allowing individuals to be personally identified, their use poses risks of personal privacy violations. In addition, because IoT-ready home appliances are connected to the Internet, camera-captured images of individual users may be unintentionally leaked. In developing our human-detection method [33], [34], we proposed techniques for detecting humans from unclear images in which individuals cannot be identified; however, a drawback of this method was its inability to detect moving humans. Thus, to enable tracking of humans even through the images are blurred to protect privacy, we introduce a particle-filter framework and propose a human-tracking method based on motion detection and heart-rate detection. We also show how the use of integral images [32] can accelerate the execution of our algorithms. In performance tests involving unclear images, the proposed method yields results superior to those obtained with the existing mean-shift method or with a face-detection method based on Haar-like features. We confirm the acceleration afforded by the use of integral images and show that the speed of our method is sufficient to enable real-time operation. Moreover, we demonstrate that the proposed method allows successful tracking even in cases where the posture of the individual changes, such as when the person lies down, a situation that arises in real-world usage environments. We discuss the reasons behind the superior behavior of our method in performance tests compared to those of other methods.
URL: https://global.ieice.org/en_transactions/communications/10.1587/transcom.2018SEP0006/_p
Copiar
@ARTICLE{e102-b_4_708,
author={Toshihiro KITAJIMA, Edwardo Arata Y. MURAKAMI, Shunsuke YOSHIMOTO, Yoshihiro KURODA, Osamu OSHIRO, },
journal={IEICE TRANSACTIONS on Communications},
title={Privacy-Aware Human-Detection and Tracking System Using Biological Signals},
year={2019},
volume={E102-B},
number={4},
pages={708-721},
abstract={The arrival of the era of the Internet of Things (IoT) has ensured the ubiquity of human-sensing technologies. Cameras have become inexpensive instruments for human sensing and have been increasingly used for this purpose. Because cameras produce large quantities of information, they are powerful tools for sensing; however, because camera images contain information allowing individuals to be personally identified, their use poses risks of personal privacy violations. In addition, because IoT-ready home appliances are connected to the Internet, camera-captured images of individual users may be unintentionally leaked. In developing our human-detection method [33], [34], we proposed techniques for detecting humans from unclear images in which individuals cannot be identified; however, a drawback of this method was its inability to detect moving humans. Thus, to enable tracking of humans even through the images are blurred to protect privacy, we introduce a particle-filter framework and propose a human-tracking method based on motion detection and heart-rate detection. We also show how the use of integral images [32] can accelerate the execution of our algorithms. In performance tests involving unclear images, the proposed method yields results superior to those obtained with the existing mean-shift method or with a face-detection method based on Haar-like features. We confirm the acceleration afforded by the use of integral images and show that the speed of our method is sufficient to enable real-time operation. Moreover, we demonstrate that the proposed method allows successful tracking even in cases where the posture of the individual changes, such as when the person lies down, a situation that arises in real-world usage environments. We discuss the reasons behind the superior behavior of our method in performance tests compared to those of other methods.},
keywords={},
doi={10.1587/transcom.2018SEP0006},
ISSN={1745-1345},
month={April},}
Copiar
TY - JOUR
TI - Privacy-Aware Human-Detection and Tracking System Using Biological Signals
T2 - IEICE TRANSACTIONS on Communications
SP - 708
EP - 721
AU - Toshihiro KITAJIMA
AU - Edwardo Arata Y. MURAKAMI
AU - Shunsuke YOSHIMOTO
AU - Yoshihiro KURODA
AU - Osamu OSHIRO
PY - 2019
DO - 10.1587/transcom.2018SEP0006
JO - IEICE TRANSACTIONS on Communications
SN - 1745-1345
VL - E102-B
IS - 4
JA - IEICE TRANSACTIONS on Communications
Y1 - April 2019
AB - The arrival of the era of the Internet of Things (IoT) has ensured the ubiquity of human-sensing technologies. Cameras have become inexpensive instruments for human sensing and have been increasingly used for this purpose. Because cameras produce large quantities of information, they are powerful tools for sensing; however, because camera images contain information allowing individuals to be personally identified, their use poses risks of personal privacy violations. In addition, because IoT-ready home appliances are connected to the Internet, camera-captured images of individual users may be unintentionally leaked. In developing our human-detection method [33], [34], we proposed techniques for detecting humans from unclear images in which individuals cannot be identified; however, a drawback of this method was its inability to detect moving humans. Thus, to enable tracking of humans even through the images are blurred to protect privacy, we introduce a particle-filter framework and propose a human-tracking method based on motion detection and heart-rate detection. We also show how the use of integral images [32] can accelerate the execution of our algorithms. In performance tests involving unclear images, the proposed method yields results superior to those obtained with the existing mean-shift method or with a face-detection method based on Haar-like features. We confirm the acceleration afforded by the use of integral images and show that the speed of our method is sufficient to enable real-time operation. Moreover, we demonstrate that the proposed method allows successful tracking even in cases where the posture of the individual changes, such as when the person lies down, a situation that arises in real-world usage environments. We discuss the reasons behind the superior behavior of our method in performance tests compared to those of other methods.
ER -