Implementation of Denoising MEMS Accelerometer Using Wavelet Transformation

Abstract

The MEMS accelerometer is an important sensor to measure the object’s motion state (acceleration), and it is a key device in inertial systems such as, vibration measurement and navigation systems. MEMS accelerometer has the advantages of high precision and large measurement range, but it suffers from noise problems which results in negative consequences, because of double integration process to calculate the distance which results in erroneous measurements especially when integrated in IMU or INS systems, that may jeopardize safety. In this research wavelet transformation method evaluated against Kalman and Kalman-wavelet combination methods, multi test scenarios were examined to select the outperformed method to be implemented using MPU-6050 sensor, and Arduino DUE development board. From the above mentioned multi test scenarios it concludes that the developed Wavelet transformation-based denoising algorithm outperforms Kalman and Kalman-Wavelet combination filtering methods by a percentage of 82% and 80% respectively so that it was the optimum method to be implemented as hypothesized in this research. Finally it was implemented successfully in the single microcontroller platform (Arduino DUE) but with certain limitation in the decomposition level and the type of the mother wavelet used in addition to the number of signal samples.