Gait Data for Efficient Gender Recognition

Abstract

Biometric recognition applications have been frequently used nowadays mostly because of reliability and ease of use for automated detection. There are many applications based on eyes, face, fingerprint, and voice for authentication and gender classification. In this paper, we focused on gender detection using the features of the steps of people. A different biometric sign has been investigated. Gait analyses were examined to determine the gender information of the people. Basic parameters like speed, variability, and symmetry of a gait, its several temporary, spatial, and height parameters, which were obtained via Physilog 5 sensor, were used in the analysis. A 321-D feature vector was comprised based on these features and an Artificial Neural Networks (ANN) model was trained with them. 95.83% accuracy was obtained. The experimental results show the success of the proposed ANN-based gait analysis system against the stateof-the-art for gender classification.