Non-Immersive method of Accelerometer calibration for mobile devices

This paper describes a method of calibrating tri-axial accelerometer sensor without the need of controlled environment and user intervention, available in present day smartphones. These sensors have misalignment errors which are induced during soldering of chip on PCB. Resulting imperfections in accelerometer sensor such as scaling factors and offsets affect the output of the sensor leading to erroneous performance of applications such as navigation, screen rotation, games etc. The paper describes a calibration model which incorporates the fundamental principle i.e. the norm of the sensor output is equal to the earth gravity acceleration. The calibration method evaluates the offset and scale factor for each axis which are computed using a non-linear minimization technique. The method is an iterative process of minimizing the mean square error. The proposed technique ensures that the data points collected doesn't have acceleration due to movement. Post calculation, verification of the output i.e. offset and scaling factor is done to ensure that the correct offsets are written in a persistent file system. The resulting algorithm is device independent and can be used in multiple platforms. Simulation and experimental results shows that the calibration output is unerring and significant improvement in performance of applications using accelerometer sensors.