Gait Abnormality Detection without Clinical Intervention Using Wearable Sensors and Machine Learning

The gait pattern varies from person to person. However, the gait of a normal healthy human differs substantially from that of an individual with an abnormal gait. The gait abnormalities could arise from various underlying health conditions such as rheumatoid arthritis, injuries, etc. The gaits of individuals could be studied to differentiate healthy gaits from abnormal gaits which can help to identify underline health conditions and for preventive medication. In this work, for the identification of abnormal gait, we have used wearable sensors, comprising of tri-axial accelerometer and tri-axial gyroscope that track’s the motion signatures produced while walking. We have collected such motion signatures of healthy persons and persons with walking abnormalities over a period of time. Later performed classification on collected data using different machine learning algorithms to segregate the cases with abnormality which can be used for facilitating remote detection of gait abnormalities using wearable sensors. With our approach, we have successfully classified different cases with an accuracy of over 90% . In the future, implementing such models in the real-world scenario could be very beneficial for the detection of gait diseases remotely with the help of smart devices like smartphones without clinical intervention.