Adaptive sensor fusion technology for mobile and wearable applications

Sensor fusion algorithms convert multiple motion sensor data such as accelerometer, gyroscope and magnetometer to generate device orientation information in terms of quaternion. Application developers also require orientation estimation capabilities on low cost devices which may have fewer motion sensor types compared to high end devices. This paper introduces an adaptive approach where application developers could configure the number of sensors and sampling rate to control the sensor fusion algorithm to achieve rotation quaternion of the required accuracy and power consumption in an efficient manner. This paper also introduces a method to generate orientation quaternion, after compensating for drift and noise, using only the low power motion sensors accelerometer and gyroscope. The comparison of the power consumed by the two sensor input fusion compared to a three sensor input fusion shows an improvement of 300%.