Enhancing the robustness of forensic gait analysis against near-distance viewing direction differences

Abstract

Gait analysis is a promising biometric technology to visually and quantitatively analyze an individual’s walking style. In Japan, silhouette-based quantitative gait analyses have been implemented as a forensic tool; however, several challenges remain owing the narrow range of application. One of the yet-unsolved issues pertains to the existence of a ‘slight’ but critical viewing direction difference, which leads to the incorrect judgment in the analyses of a person even when using deep learning-based feature extraction. To alleviate the critical viewing direction difference problem, we developed a novel gait analysis technique involving three components: 3D calibration, gait energy image space registration, and regression of the distance vector. Results of the GUI development and mock appraisal tests indicated that the proposed method can help achieve practical improvements in the forensic science domain.

Interpol review of forensic video analysis, 2019–2022

“…The standard of evidence admissibility in the United States also has been discussed [ 104 , 105 ] Furthermore, there are many papers that discuss the reliability of gait analysis [ 60 , 68 , 72 , 89 , 93 , 94 , 96 , 105 , 106 , 108 , 116 , 120 , 123 , 125 , 126 , 129 , 130 , 132 , 136 , 140 , 144 ].…”

Pre‐set estimation‐based in‐silico silhouette‐based methodology for improving the robustness to viewing direction difference for assisting forensic gait analysis

“…This makes it possible to determine the same person more accurately than in [22] by considering an appropriate viewing direction for each footage. The authors devised a spatial alignment process for gait features and used a practical dataset to evaluate the accuracy [23], which successfully improved the practicality by deepening the calibration‐based gait analysis. This method can perform a more accurate analysis but involves an elaborate procedure for camera calibration, which requires considerable time for performing reverse projection and on‐site measurement [23].…”

“…However, its application in courts in the forensic domain is currently limited as several challenges remain to be addressed, and it is therefore currently only utilized for cases with ideal conditions [20, 21]. Therefore, this study uses quantitative semiautomatic silhouette‐based methodologies to investigate the problem of comparing pedestrian footage to identify whether the footages show the same person in a forensic application [22–25]. In the forensic science domain in Japan, a quantitative semiautomatic silhouette‐based method [22, 23] is used by experts to assist scientific appraisal for objectivity as evidence in court.…”

“…Therefore, this study uses quantitative semiautomatic silhouette‐based methodologies to investigate the problem of comparing pedestrian footage to identify whether the footages show the same person in a forensic application [22–25]. In the forensic science domain in Japan, a quantitative semiautomatic silhouette‐based method [22, 23] is used by experts to assist scientific appraisal for objectivity as evidence in court. Even if lens distortion is corrected, as in [22] wherein a pre‐set approximation from 24 conditions is performed, different viewing directions can become a very critical problem for conducting a correct analysis of the same person; that is, the false rejection rate of the same person increases significantly if the viewing directions change even slightly.…”

“…Even if lens distortion is corrected, as in [22] wherein a pre‐set approximation from 24 conditions is performed, different viewing directions can become a very critical problem for conducting a correct analysis of the same person; that is, the false rejection rate of the same person increases significantly if the viewing directions change even slightly. To address this problem, recently, researchers have revisited and improved the 3D camera calibration‐based gait analysis approach [16, 26] and found that the problem can be solved by performing accurate 3D camera calibration at the camera installation site [23]. Here, the viewing direction pertains to the relationship between the position and orientation of the camera regarding the pedestrian.…”