Human Gait Identification System Based on Transfer Learning

Biometric identification systems have recently made exponential advancements in term of complexity and accuracy in recognition for security purposes. Surveillance areas face and deal with several security challenges. Gait analysis is a method for human identification through biometric characteristics in human locomotion. The objective of this project is to propose an advanced and accurate end-user software system that is able to identify people in video based according to their gait signature for hospital security purposes. Transfer learning based on a pre-trained Convolutional Neural Network has been used. It is able to extract deep feature vectors and classify people directly instead of traditional representations that include computing the binary silhouettes and hand-crafted feature engineering. The results indicate that the training and testing accuracies of the proposed approach were 100
^%
and 93.57% respectively. As a conclusion, this implemented biometric system outperforms the traditional neural network approaches in gait recognition that require multiple parameter tuning and that face difficulty in data training. It can also be considered as an effective tool in securing healthcare field.