Classification of human moving patterns using air pressure and acceleration

Abstract

A new method to classify human moving behavior using one acceleration sensor and one air pressure sensor is discussed. Foot-to-head acceleration frequently varies during walking, and is used to classify the moving behavior into stepping action and rest. Walking patterns such as normal walking and jogging are estimated from variances of foot-to-head acceleration. Slight change of air pressure which results from vertical movement is used to detect the moving styles of going up/down the stairs or in an elevator. A combination of these features gives several kinds of moving patterns. The proposed method enables simple and reliable estimation of human moving pattern at the rate of 95% accuracy.