Abstract
Human gait is a pattern of biometric movement for human identification. Unlike other biometrics such as fingerprint, iris, face, and voice recognition, human gait can be captured with unobtrusive method. In this paper, modeling of human body for gait recognition using depth camera is proposed. Using the depth data, modeling gait features based on model-based method can be extracted and measured. The raw depth data is captured in 3D space from the Kinect device. Kinect 2.0 is based on ToF method for depth information. Thus, two cameras are used for capturing a walking person on sides to reduce errors by obstacles in the light from the camera. Each depth data is required to capture range setup, segmentation of body parts, and identification of body joints using our proposed method. Then the body joints are calculated for three-angles (hip/knee/ankle) in a human gait cycle. The angles using the reconstructed body joints are more stable and accurate than using the Kinect skeleton method for human gait analysis.