Skip to Main Content
These days smart phones have changed the concept of mobile platform. Due to rapid growth in affordability, increased sensory and computational power has opened the ways for some interesting applications like activity recognition, gaming, navigation/tracking, augmented reality etc. In all these applications orientation of mobile phone is very important, once we know the correct orientation of the device, it becomes easy to develop such applications. Orientation can be determined by sensor fusion of accelerometer and magnetometer but it provides good accuracy as long as device is stationary or not moving linearly and also it suffers from surrounding magnetic interference. Orientation estimation systems often use gyroscope to increase reliability and accuracy. Although gyroscopes provide a quick response to change in angles and do not have problems like interference, they suffer from bias and integration errors which introduce drift in signal. In this paper an efficient and less complicated approach is utilized which minimizes the drift and noise in output orientation. In most of smart phone navigation applications these days, it is assumed that mobile is at fixed position like holding in hand facing forward, in trouser/jacket pocket or attached to waist. But if we assume that mobile position is changing arbitrary then situation becomes complicated. The goal of this work is to find accurately the direction of movement of a pedestrian while holding mobile in any position. In this work an accelerometer based approach is presented for pedestrian direction of movement estimation.