Gait Recognition using Segmented Motion Flow Energy Image

Gait recognition is becoming an apparent form of biometric recognition method which dominates the area of unobtrusive biometric recognition. Many methods have been established for gait identification which give a fairly accurate recognition rate. Gait recognition is very sensitive to appearance changes of subject. The approach described in this paper tries to solve the major challenge of gait recognition which is to deal with covariates such as walking speed, clothing and carrying bag. It is very essential to choose the most proper gait features that highlight the special trait of the subject. Motion images are calculated by diminution of previous adjacent gait silhouette. Then the final feature image is calculated by segmentation of Motion Flow Energy Image (MFEI) which is the average of motion images calculated from same procedure as of GEI. After this, the extracted feature image is used for training and testing. The classifiers Principal Component Analysis along with Linear Discriminant Analysis (PCA+LDA) and Linear Support Vector Classifier gives a substantial accuracy under covariate conditions. The experiments are performed on CASIA-B and CASIA-C datasets. The experimental results on the datasets signify that the proposed method gives remarkable results.