Gait Analysis using 6DoF Magnetic Tracking

Abstract

Gait analysis using wearable devices has advantages over non-wearable devices when it comes to portability and accessibility. However, non-wearable devices have consistently shown superior performance in terms of the gait information they can provide. This calls for the need to improve the performance of wearable device based gait analysis. To that end, we developed a 6 Degrees-of-Freedom (6DoF) magnetic tracking based gait analysis system as a step in this direction. The system is portable, minimally intrusive, wireless and power efficient. As a proof-of-concept, the system was used for the task of Human Activity Recognition (HAR) to classify four tasks - walking (W), walking with weight (WW), jogging (J) and marching on the spot (M). Gait data of 12 participants was collected. The classification performance of two deep learning (DL) classifiers - Convolutional Neural Networks (CNN) and Long Short Term Memory (LSTM) - was compared. The performance of the magnetic tracking based gait analysis system was also compared with an Inertial Measurement Unit (IMU) + magnetometer based system. The magnetic tracking based system showed an overall classification accuracy of 92\% compared to 86.69\% for the IMU + magnetometer system. Moreover, the magnetic tracking system showed an improvement of about 8\% in being able to differentiate between W and WW. This highlights the insufficiency in the information content in the data from IMU + magnetometer, warranting the need for a complete 6DoF tracking. Our work, thus, proves the feasibility of using magnetic tracking systems for the purpose of gait analysis.