Human joints Analysis System: A Machine Learning Approach

Abstract

Bipedal robots are the humanoid robots and now a days they are the gems of industries as well as other fields. Still bipedal robots are suffering from sort of locomotion problems as they could not perfectly adopted the natural gait of human beings. Therefore, still researchers are working on collecting human gait data, analysing those data and making bipedal robot to adopt that gait patterns. Researchers are struggling to optimize an appropriate gait pattern which is suitable to walk like human, can handle external impact of surrounding, can deal with the abnormal surface and still can maintain balance during walk. In this research article, a machine learning based gait pattern identification mechanism is suggested. To address the aforementioned task, well known machine learning models are opted, namely, Logistic Regression, Random Forest, Linear SVM individually and then Hybrid model (combination of logistic regression, random forest, and liner SVM). In the obtained performance analysis, it is observed that suggested hybrid machine learning model improves the accuracy of gait pattern, therefore, further these results can be opted in bipedal robots for an effective and efficient walk. In the performance measure, hybrid model achieved 98.57% accuracy.