Extraction of bodily features for gait recognition and gait attractiveness evaluation

Abstract

Although there has been much previous research on which bodily features are most important in gait analysis, the questions of which features should be extracted from gait, and why these features in particular should be extracted, have not been convincingly answered. The primary goal of the study reported here was to take an analytical approach to answering these questions, in the context of identifying the features that are most important for gait recognition and gait attractiveness evaluation. Using precise 3D gait motion data obtained from motion capture, we analyzed the relative motions from different body segments to a root marker (located on the lower back) of 30 males by the fixed root method, and compared them with the original motions without fixing root. Some particular features were obtained by principal component analysis (PCA). The left lower arm, lower legs and hips were identified as important features for gait recognition. For gait attractiveness evaluation, the lower legs were recognized as important features.

SkeletonGait: Gait Recognition Using Skeleton Maps

“…Additionally, there are also some progressive multi-modal gait frameworks, such as SMPLGait (Zheng et al 2022) that exploited the 3D geometrical information from the SMPL model to enhance the gait appearance feature learning, and BiFusion (Peng et al 2023) that integrated skeletons and silhouettes to capture the rich gait spatiotemporal features. Related Works to Skeleton Map.…”