Personalized Knee Angle Prediction Models Using Machine Learning

Abstract

Gait analysis had been traditionally used to diagnose underlying pathological conditions, but recently it has seen widespread applications in varied fields like bio-metrics, rehabilitation, sports, animation, etc. This study focuses on the rehabilitation prospects of lower limb amputees and to accurately predict their natural knee angle using easily available body parameters. This would ensure easier and better rehabilitation. The subjects included in the study belong to the MNIT Gait Dataset, collected by RAMAN Lab in MNIT Jaipur. For analysis, the study compares various supervised machine learning models across several regression evaluation metrics to achieve the final objective of predicting a subject’s knee angle accurately. The results from this study can be used in areas with low technology penetration for better patient rehabilitation.