Classification of Human Gender Through Machine Learning Analysis of Gait

The study of human gait patterns can provide valuable information, such as identifying an individual’s characteristics and preferences. This paper aims to determine whether it is possible to use machine learning algorithms to classify one of the fundamental human features, gender, based on joint angles measured from gait. This study is limited to only the male and female genders. Pose estimation by MediaPipe is used to build the dataset in this study. The performance of four machine learning algorithms: Support Vector Machines, Gaussian Naive Bayes, Random Forest, and XGBoost are compared. The results show that XGBoost model achieved the best classification accuracy of 97%. This study hopes to contribute in the ongoing research on human gait characteristics.