Benchmarking the accuracy of inertial sensors in cell phones.

摘要

Many ubiquitous computing applications rely on data from a cell phone's inertial sensors. Unfortunately, the accuracy of this data is often unknown, which impedes predictive analysis of applications that require high sensor accuracy (e.g., dead reckoning). This work focuses on benchmarking the accuracy of the accelerometers and gyroscopes on a cell phone. The cell phones are attached to a robotic arm, which provides ground truth measurements. The misalignment between the cell phone's and the robotic arm's reference systems is computed using Horn's algorithm for closed-form absolute orientation estimation. Two cell phones (Apple's iPhone 4 and Nokia's N97) have been tested, and results are provided in terms of random noise, error bias, error bias correlations, and orientation reconstruction error.