A Framework for Silhouette Inspired Robust GAIT Based Human Identification | Original Article

ABSTRACT:

Gait has enticed substantial consideration in the current years as it has number of unique features that are not present in other biometrics techniques. The features like videos of subjects can be used which are recorded from a distance that is without alerting the subject at all and secondly, videos recorded at low resolution can be used for recognition. The researchers have shown interest in gait as its recognition results have shown a promising growth in the controlled environment. In this thesis, solutions are proposed for the real world by taking care of gait controlled situations like view conditions, resolution and fluctuation of gait patterns due to carrying clothes as well as goods. To achieve the human identification in a better way, a new approach has been used. Erosion method for cleaning operation has been used that is followed by the convolution neural network with two convolution layers and pooling layers have been used to optimize the human recognition system. All the results have been obtained from CASIA dataset as well as OU-ISIR data set.