부산대학교 도서관

Gait is a representative of an individual walking or a walking pattern. The definition of the gait cycle focuses on the lower limbs. However, the whole body has motion when humans walk. If we understand more about the gait features, it will benefit the gait analysis pipeline. This paper presents the walking pattern matching based on human body parts by calculating the angles between joints (the body angles) from vision-based skeletal joint coordinates. Then, apply Dynamic Time Warping (DTW) to demonstrate the similarity between persons by matching their whole body, upper body, and lower body feature vectors. To study the effect of carrying and clothes variations on the walking pattern when applying DTW matching. The authors aim to study human walking behavior by employing this finding to create a suitable gait image analysis based on a human body parts model to improve a computer-based gait analysis.