An Adaptive Smoothing Method for Sensor Noise in Augmented Reality Applications on Smartphones

Handling inaccurate and noisy sensor readings are among important challenges while implementing augmented reality applications on smartphones. As a result, we need to smooth the sensor readings for steady operation. However, no smoothing algorithm performs best in all cases as there is an inherent tradeoff. On one hand, excessive smoothing slows down the effect of device movements, hence makes applications less responsive. On the other hand, insufficient smoothing causes objects on the screen to constantly move back and forth even while the device is steady, hence makes applications too responsive. Clearly, both of the extremes cause augmented reality applications to be less effective in terms of human-computer interaction performance. In this paper, we propose an adaptive smoothing method based on the rate of change in device view direction. Basically, the method adjusts the smoothing level adaptively based on the phone movement. Our experimental results show that our adaptive approach, in comparison to previous proposals, achieves a better smoothing for various cases of phone movements.