On Stair Walk Recognition Using a Single Magnetometer-free IMU and Deep Learning

Abstract Human activity recognition (HAR) is an expanding area of research. Although sensors are becoming more readily available, there is a trend toward minimizing the number of associated sensors for better applicability. Neural networks are often employed for HAR, as they can identify movement patterns within data. To optimize the amount of useful information provided to the network, feature extraction methods are commonly applied. However, these feature methods complicate the data processing pipeline, and thus are less applicable to continuous real-time applications. In this work, we investigate the impact of using quaternions as input features for an LSTM network on HAR, specifically focusing on level walking and stair climbing activities, while also considering the inference time. We demonstrate that combining quaternions and raw IMU data, i. e., acceleration and angular rates significantly enhances classification accuracy without adversely affecting the inference time. By additionally adding an approximate estimate of the vertical position change to the input data, the classification accuracy is further improved to a value of 96.18%.