Development of a Novel Recognition Method of Model-based Gait Recognition based on the Mechanism of Walking

2016 Fiscal Year Final Research Report

• Project/Area Number: 15H06874
• Research Category: Grant-in-Aid for Research Activity Start-up
• Allocation Type: Single-year Grants
• Research Field: Intelligent mechanics/Mechanical systems
• Research Institution: National Research Institute of Police Science
• Principal Investigator: Imoto Daisuke 科学警察研究所, 法科学第二部, 研究員 (10760902)
• Project Period (FY): 2015-08-28 – 2017-03-31
• Keywords: 歩容認証法 / モデルベース手法 / 関節位置の確率分布 / 非線形数理モデル

Outline of Final Research Achievements

Model-based Gait Recognition is a method by which we can recognize two pedestrians in CCTV videos is a same person or are different people from dynamics of joints’ coordinates (Conventional method). It is a problem that intra-individual variability, for example different clothes, sometimes deteriorate the accuracy of the analysis. Here we propose two methods which is based on the mechanism of walking. One is a method using estimated parameters of coupled nonlinear oscillators’ model. The other is a method using dynamics of probability densities computed from presumably learned spatial distributions of joints’ positions. It is revealed that the method using coupled nonlinear oscillators’ are not useful for forensic purpose, on the other hand the method via spatial joints’ distributions may be useful in a different clothes’ case. It is suggested that representation of walking using spatial joints’ distributions could be overcome a part of intra-individual variability.

Free Research Field

法科学、情報学、数理生物学