Gait Feature Analysis for Personal De-identification

This paper aims for analyzing gait to investigate feature values related to gait transformation using 3D(Dimensional) human data in order to conceal a personal identification. Biometric data including fingerprint, iris, vein, and gait etc., are distinctive characteristics that certificate individuals. Gait is useful even if it is taken from faraway place. A Kinect is used to obtain gait data. 3D coordinates of both elbows, both knees and both wrists are measured. Thirty gait features are calculated by these coordinates. Moreover, these sequential data are transformed into frequency domain by Fourier transform. Individual motion periods is analyzed by using upper fourth power spectrum, and the obtained period is defined. Obtained thirty one gait data are analyzed by principal component analysis. As a result, both Wrists heights, left knee angle and right elbow angle are the significant factors on a human gait feature.