CLASSIFICATION OF HUMAN GAIT ACCELERATION DATA USING CONVOLUTIONAL NEURAL NETWORKS

The human motion analysis using wearable sensors such as accelerometers and gyroscopes is one of the important issues in ubiquitous and wearable computing. Inspired by a paper by Akiduki et al. that was released in 2018 concerning the classification of human gait motion accelerometer data, this paper attempts to classify that same data using a convolutional neural network. In the original 2018 paper, a high degree of separation was found between the data of the 13 recorded test subjects, suggesting that classification purely by looking at the motion data is possible. For the purpose of classification using the neural network, the given time series data is converted into three matrices (equivalent to image data with three channels per pixel). Using these images as input for a convolutional neural network, an accuracy of 100% was achieved in classifying the subject number from previously unseen motion data.