紀要論文

歩行周期間における空間ゆらぎ画像列を用いた歩行者年齢推定

Gait-based Age Estimation using Spatial Fluctuating Image Sequences during the Gait Cycle

jpn

24321192

https://doi.org/10.15002/00030610

法政大学大学院情報科学研究科

法政大学大学院紀要. 情報科学研究科編

19

1

6

2024-03-24

Version of Record

Since gait contains personal characteristics, it has been widely studied and put into practical use as a biometric authentication modality. Gait recognition is a technique based on image recognition and uses the walking silhouette of the subject. However, waking is not a perfectly periodic movement, the observed silhouettes can fluctuates depending on a cycle. From the perspective of physical therapy and gait analysis, we hypothesized that this fluctuation is related to changes in gait with growth and aging. And that the fluctuation could be utilized as a feature in age estimation. In this study, we propose to use spatial fluctuations calculated from multiple gait cycles for age estimation. Spatial fluctuations are represented as a sequence of differential silhouette images associated with two gait cycles. The feature extractor trained on these differential silhouette image sequences is used to estimate actual age. We report the evaluation results. As a result, the mean absolute error (MAE) using three shooting angles showed an improvement of approximately 0.1 years in estimation error. When only the 0° shooting angle was used for training, an improvement of approximately 0.5 years in estimation error was observed.

Article

AA12746425