Prediction of Gender and Age from Inertial Sensor-based Gait Dataset

The technology pertaining to gait generation, recognition and analysis is getting more sophisticated each day. The global demand for a gait-based dataset and its subsequent recognition, leading to extraction of valuable information is now higher than ever. However, inertial sensor-based gait dataset is a comparatively new addition to the field of gait analysis. Consequently, most of the research works incorporating machine learning algorithms into gait dataset are image based. In addition to that, most of the gait-based datasets have been analyzed for gait recognition. There remains very little research work on personal authentication from inertial sensor-based gait dataset. Personal authentication is of several types out of which, predicting gender and age is quite challenging. In this paper, we have tried to face these challenges and have manifested the process of predicting gender and age from the inertial sensor-based gait dataset which is a part of the vast Osaka University-ISIR Gait Database. Finally, we have found that the models, Support Vector Machine shows the highest accuracy for the problem of classifying gender and Decision Trees shows the highest variance score (R
²
value) for the problem of predicting age.