Estimation of Walking Speed by Signal Energy of Frequency Band of Walking Acceleration

Abstract

This paper describes a new method of walking speed estimation by a simple accelerometry of ambulatory movements, aimed at a precise quantification of energy consumption. It was postulated that walking speed variation reflects both strength and frequency of pelvis vibration during walking. Therefore, amplitudes of vertical acceleration that come along with walking speed variation were evaluated in terms of the signal energy in walking frequency band 1-3Hz. According to the Parseval's Theorem, signal energy can be applied both in time domain and frequency domain signal processing. The relationship between the walking speed and the signal energy was investigated considering aging. A portable instrument attaching on the waist measured vertical acceleration. Then, an estimate equation of walking speed as a function of signal energy was formulated from the experimental results of 199 subjects including young and elderly people. To validate the estimation, another 85 elderly people including 5 subjects who walked with the aid of sticks were investigated. The subjects were instructed to walk on a 16-meter strait level track at self-selected normal walking speed individually. Walking speed was estimated for 10 second continuous walking. Lap time was also measured to evaluate real walking speed. Results showed that the walking speed was accurately estimated by the proposed method (95%). It was also confirmed that this method could be applied to estimate the walking speed of both young and elderly people, but not those who walk with the aid of sticks.