Temporal Characteristics of Gait Biometrics

This paper presents a gait-based user recognition and also provides evidence that recognition performance can decrease, when the time elapsed between template and test samples of user's gait is lengthy. Nevertheless, deteriorated performance of the gait recognition system is still statistically significant. Unlike most of other gait recognition methods which use video source (silhouettes), our approach is based on acceleration data of person's normal walking. Acceleration data is obtained using a sensor, which is attached to person's leg. On two separate sessions 74 gait sequences of normal walking from 10 subjects are collected. On the same day comparisons best correct classification rate (CCR) is 90% while on the cross day comparisons best CCR is 59%. On cross day comparison the time gap between template and test sets was 16 days on average.