Recognition of walking behaviors for pedestrian navigation

Abstract

This paper presents a method for detecting and classifying walking behaviors based on acceleration measurements of a pedestrian, and is employed in an indoor navigation system currently being developed. The prototype navigation system uses a set of inexpensive and wearable sensors: a bi-axial accelerometer, a digital compass, and an infrared light detector. Using the measured acceleration data, the proposed method can detect forward steps and classify the steps as: "level ground", "up", and "down". The objective of the detection is to count steps for estimating the current position by dead-reckoning using heading measurements. The capability in detecting "up/down" steps can be used to correct estimated position errors. The effectiveness of the proposed method is demonstrated by experiments on six persons.