μIMU-based handwriting recognition calibration by optical tracking

Abstract

A MAG-mulMU which is based on MEMS accelerometers, gyroscopes and magnetometers has been developed for real-time estimation of human hand motions. Appropriate filtering, transformation and sensor fusion techniques are combined in the ubiquitous digital writing instrument (UDWI) to record the handwriting on any surface. However, because of the sensors' intrinsic biases and random noise such as circuit thermal noise, a calibration system that provides good reference measurement parameters must be used to compare the output of the MAG-mulMU sensors. A basic concept to calibrate three-dimensional linear accelerations, angular velocities and space attitude through optical tracking techniques was proposed and proved feasible in our previous paper. The optical tracking system (OTS) developed by our group consists of two parts: (1) 2D trajectory calibration that is used to obtain linear accelerations of the UDWI in a particular frame defined by us; (2) multiple camera calibration that is used for attitude calibration of the UDWI. An essential relationship to transform reference frames and angular velocities can be guaranteed after real-time attitude calibration. Some improvements have been adopted in part 1 and the latest experimental results about the calibration of the accelerometers will be introduced and discussed in this paper.