Gait analysis and identification

An efficient framework were proposed for identifying individuals from gait via feature-based method based on 3D motion capture data. Three different extraction methods were applied to achieve gait signatures. The average identification rate was over 93% with best result close to 100% in a 35 subject database. In additional, gait attractiveness was analyzed via Principle component analysis and linear regression method. A systematic relationship was found between the motions of individual markers and the attractiveness ratings. In a linear equation, ln(PCA1) and ln(PCA2) predicted ln(attract_value) with reasonable accuracy.