Abstract—For mobile applications such as command interfaces and canine healthcare, we propose a method that can detect rotation by analyzing the motion locus calculated from motion sensor data such as three axis angular velocities. While conventional methods demand the use of multiple sensors, our proposal uses only a single gyro-sensor. The method judges rotation by setting criteria based on the distortion of normal vectors extracted from the motion locus. The distortion of normal vectors is defined by principle component analysis. Therefore, it has low computation costs. We evaluate two variants of the proposed method in a comparison of normal vector extraction and confirm their superior performance. The results show the method can be easily implemented on not so high-spec mobile devices.
Index Terms—Principal component analysis, rotational motion, motion locus, cross product, plane approximation.
Toshiki Iso is with Research Laboratories, NTT DOCOMO, INC. Kanagawa, Japan (e-mail: isot@nttdocomo.com).
[PDF]
Cite: Toshiki Iso, "Motion Locus Analysis to Detect Rotation," International Journal of Information and Electronics Engineering vol. 7, no. 6, pp. 147-151, 2017.