A recognition of human walking motion by using an accelerometer

Abstract

This paper presents a recognize and visualize methods of human motion patterns in walking using an accelerometer for automatic emergency call systems for seniors. The recognition is realized by these two steps: (1) extraction of a spectrum of the acceleration measured by a accelerometer attached on the waist that represents the gait cycle (cycle of foot steps) through dynamic threshold processing; (2) classification of the motion based on the observed fact that the intensity of the spectrum increases when the human motion changes from 'stand', 'walk' to 'run'. Moreover, the result of the recognition is visualized as a real-time, privacy-protected animation, where a character animation generator, our previous study work, is used that a character is controlled by qualified motion parameters.