Drift modeling and compensation for MEMS-based gyroscope using a Wiener-type recurrent neural network

This paper presents a self-designed wearable inertial sensing module and its associated drift modeling and compensation algorithm to obtain drift-removed signals for handwriting trajectory reconstruction. Users can utilize the wearable inertial sensing module to write numerals at normal speed, which consists of a microcontroller, a triaxial accelerometer, a uniaxial gyroscope, a biaxial gyroscope, and an RF wireless transmission module. The accelerations and angular velocities generated by the handwriting movements are transmitted to a computer via the RF module. In order to minimize the integral or cumulative errors caused by the intrinsic drift of the microelectro-mechanical-system (MEMS)-based gyroscopes, we have developed a drift modeling and compensation algorithm for removing the drift of the gyroscopes and increasing the accuracy of trajectory reconstruction. In the proposed drift modeling and compensation algorithm, a Wiener-type recurrent neural network (WRNN), which consists of a dynamic linear model cascaded by a static nonlinear model and can be expressed in a state-space representation, is used to model the drift of the gyroscopes. Our experimental results on handwriting trajectory reconstruction have successfully validated the effectiveness of the self-designed wearable inertial sensing module and its drift modeling and compensation algorithm.