A Noise Reduction Method for IMU and Its Application on Handwriting Trajectory Reconstruction

Abstract

In this paper we purpose to develop a digital pen based on low-cost IMU (Inertial Measurement Unit) There are some advantages of using IMU to design a digital pen: 1) Unlimited writing space 2) portable and 3) power saving The IMU used in our digital pen contains a 3-axis accelerometer and a 3-axis gyroscope The accelerometer is utilized to measure the acceleration when the digital pen is moving and we can know how the digital pen moves based on acceleration information The gyroscope can measure the rotation of the digital pen and the rotation information is used to improve the precision of acceleration However there are intrinsic bias and random noise in the IMU signals The noise maybe caused by the internal thermal mechanical fluctuations environment temperature and the user's unconscious trembles Therefore we propose a handwriting trajectory reconstruction method which can be applied to IMU sensors with high-degree noise To more accurately reconstruct the trajectory we apply the machine learning technique to detect the movement state of the digital pen In previous related works they set a simple threshold to detect the pen movement which is not robust and general for all IMU sensors We extract several features from IMU signals and train a model for movement detection based on LDA (Linear Discriminant Analysis) We also propose a mechanism named "reset switch" to more effectively restrain the accumulated error of the accelerometer

Date of Award	2014 Dec 3
Original language	English
Supervisor	Meng-Hsun Tsai (Supervisor)