High-dimensional Time-series Gait Analysis using a Full-body Wireless Wearable Motion Sensing System and Convolutional Neural Network

This paper reports high-dimensional time-series data processing for gait analysis using data from a full-body wireless wearable motion sensing system with machine learning models. The wearable sensing system consists of ten sensors deployed on limb landmarks, which record quaternion data at 59 sample/Second. Two pilot clinical studies have been designed to investigate the capability of the sensors for accurately differentiating normal gait parameters in healthy individuals. The two studies are interlimb coordination correlation and gait task classification. The interlimb coordination study calculates the correlation coefficient between the left and the right joints to evaluate interlimb coordination. The gait task classification uses a five-layer convolutional neural network to test the classification accuracy of four tasks selected from the Functional Gait Assessment. Clinical experiments show that the system is able to identify the correlation coefficient between normal and synthesized abnormal gaits. The convolutional neural networks have the ability to differentiate functional gait tasks with up to 90% accuracy.