Spatio-temporal generative adversarial network for gait anonymization

• Published in: Journal of information security and applications Vol. 46; pp. 307 - 319
• Main Authors: Tieu, Ngoc-Dung T., Nguyen, Huy H., Nguyen-Son, Hoang-Quoc, Yamagishi, Junichi, Echizen, Isao
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.06.2019
• Subjects: Biometric feature; Deep learning; Gait; Gait anonymization; Security; Deep learning; Biometric feature; Gait anonymization; Gait; Security
• ISSN: 2214-2126
• DOI: 10.1016/j.jisa.2019.03.002

Gait anonymization for protecting a person’s identity against gait recognition while maintaining naturalness is a new research direction. It can be used to protect the identity of people in videos to be posted on social networks, in police videos that require redaction, and in videos obtained from surveillance systems. We have developed a spatio-temporal generative adversarial network (ST-GAN) that uses random noise synthesized in the gait distribution to generate anonymized gaits that appear natural. ST-GAN consists of a generator that uses the original gait and random noise to generate an anonymized gait and two discriminators, a spatial discriminator and a temporal discriminator, to estimate the probability that a gait is the original one and not an anonymized one. Evaluation showed that the anonymized gaits generated with the proposed method are more natural than those generated with an existing method and that the proposed method outperforms the existing method in preventing gaits from being recognized by a gait recognition system.