A Bidirectional Siamese Recurrent Neural Network for Accurate Gait Recognition Using Body Landmarks

Proma Hossain Progga, Md. Jobayer Rahman, Swapnil Biswas, Md. Shakil Ahmed, Arif Reza Anwary, Swakkhar Shatabda · 2024

Read A Bidirectional Siamese Recurrent Neural Network for Accurate Gait Recognition Using Body Landmarks by Proma Hossain Progga, Md. Jobayer Rahman, Swapnil Biswas, Md. Shakil Ahmed, Arif Reza Anwary, Swakkhar Shatabda (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-03498.txt.

Description

Proma Hossain Progga and co-authors identify people by how they walk, using body landmarks from video, in a paper from 2024. Gait is a useful biometric when faces or fingerprints are not practical. Their method pulls joint landmarks from each frame with the MediaPipe pose tool, then aligns the sequences with Procrustes analysis.

A bidirectional Siamese network with gated recurrent units learns the time pattern of each walk and compares pairs. On four large cross-view datasets, the system scored between 86.6 and 95.7 percent accuracy. The best result came on the CASIA-B set, and the hardest was the in-the-wild Gait3D set. The authors argue that landmark-based gait checks could serve real security and monitoring tasks.

Catalog record

CREATORProma Hossain Progga, Md. Jobayer Rahman, Swapnil Biswas, Md. Shakil Ahmed, Arif Reza Anwary, Swakkhar Shatabda
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE69 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-03498

Browse the full library · About the Tower of Babel