Gait pattern production based on silhouette differences

In this work, a novel gait pattern production approach is proposed for human gait recognition. Gait biometric features are determined based on silhouttes. Proposed gait pattern is obtained by sum of the different silhouttes over a gait cycle. Then transformation matrix is produced with principal component analysis of training data set obtained from gait patterns. Each pattern is transformed into another domain and classified here. Classification is done with nearest neighbor algorithm. Successful results are achieved with the proposed method even when silhouttes can't be acquired well.