Microelectromechanical system (MEMS)-based gyroscopes have been widely applied to various inertial-sensing-based human-computer interaction (HCI) devices. However, the random drift of MEMS-based gyroscopes limits their applications. Hence, studies pay attention to develop various models to model and compensate the random drift for improving the performance of the MEMS-based gyroscopes. This paper presents a self-constructing Wiener-type recurrent neural network (SCWRNN) with its false nearest-neighbors-based self-constructing strategy and recursive recurrent learning algorithm to model the random drift of the MEMS-based gyroscopes and then compensate them from the calibrated gyroscope measurement. Subsequently, the proposed random drift modeling and compensation algorithm is integrated into the handwriting trajectory reconstruction algorithm of the inertial-sensing-based HCI device, called IMUPEN, for accurately obtaining the reconstructed handwriting trajectory. Users can hold the IMUPEN, which is composed of an accelerometer, two gyroscopes, a microcontroller, and an RF wireless transmission module, to write numerals at normal speed. The accelerations and angular velocities measured by the accelerometer and gyroscopes are transmitted to a personal computer through the RF module for further reconstructing the handwriting trajectory via the handwriting trajectory reconstruction algorithm. In addition, we have developed the SCWRNN-based random drift modeling and compensation algorithm to eliminate the cumulative errors caused by the random drift of the MEMS-based gyroscopes for further increasing the accuracy of handwriting trajectory reconstruction. Our experimental results have successfully validated the effectiveness of the proposed random drift modeling and compensation algorithm and its application in handwriting trajectory reconstruction.
All Science Journal Classification (ASJC) codes
- Computer Science(all)
- Materials Science(all)