Walking movement detection using stationary stochastic methods on accelerometer data

Abstract: In the area of physical activity recognition, there is a great demand for better understanding data and building useful models for data analysis. Many studies have focused on using machine learning algorithms, which provide high accuracy but are computationally expensive. However, few studies have tried to approach this problem with statistical methods. The purpose of this study is to investigate the performance of statistical signal processing methods when applied to smartphone accelerometer data. Specifically it focuses on the distinction between walking and non-walking users, with the aim of extracting characteristics that can be useful for traffic and city planning.