Mustansiriyah Journal of Pure and Applied Sciences

People Recognition Based on Gait and Neural Network

Razan E. Al-Bayati

Ziad M. Abood

DOI: https://doi.org/10.47831/mjpas.v1i1.13

Keywords: Gait Recognition, Dynamic Gait Features, Dense Neural Network, Angle, Gait Cycle.

---

Abstract

Walking is one in every of life science that helps to see the identity of an individual while not his data and from an excellent distance during. The main objective of this study is to design cheap gait identification system by using digital camera and laptop. In this study, the proposed system consists of main stages gaiting video, 3D key joint position for each frame determine, feature extraction, cleaning features, and identification. In the first phase, Gaiting video  human body can be identified using a digital camera, Secondly, 3D key joint position for each frame determine. Thread, the features extraction phase, two methods have been used to extract features dynamics gait feature. Four, cleaning features using linear discriminant analysis. Finally, Identification by using dense neural network classifiers have been applied set of the feature. The proposed system was tested on our dataset consisting of 126 gait videos of 27 people, where the results after using 80% of the data for training and 20% for the testing showed achieving accuracy, the results of the dense classifier were for the dynamics gait feature 80%.