De-noising algorithm based on compression of wavelet coefficient for MEMS accelerometer signal

In this paper a Micro-Electro-Mechanical Systems (MEMS) accelerometer signal noise reduction method based on linear compressing of wavelet coefficient is presented. First of all, the original signal being transfered by using Mallat algorithm, a dual-threshold concept is proposed corresponding to processing the singular points and high-frequency noise. Then, the threshold for the singular point detection being calculated, through the detecting wavelet coefficients modulus maxima to determine the location and amplitude of singular points, the energy of singularity is eliminated by linear compression of neighborhood of singular points. Finally, the threshold de-noising calculating method is presented for MEMS accelerometer signal and the smooth and enhance of the signal is achieved. De-noising method using linear compression and noise reduction threshold overcome the shock phenomenon near the singular point and a better SNR is achieved than using universal threshold.