Research of de-noising for micro-mechanical gyro signal based wavelet transform

Abstract

According to a certain type of micro-machined gyroscope signal characteristics, wavelet transform de-noising method was adopted. The measured noise signal mainly concentrated in the low-frequency phase, overlapping with the gyro signal spectrum, and therefore the wavelet should have a certain degree of vanishing moments, in order to reduce the calculation and the distortion of reconstructed signal at the edges. The wavelet also has compact and symmetry properties of properties, using wavelet Symlets. According to the real signal in the wavelet decomposition of a higher layer of relatively large wavelet coefficients, the principle was determined with the variance before and after an order of magnitude difference, and calculating the decomposition level is 5. The five-scale decomposition was done with five kinds of wavelets for signal denoising experiments, and de-noising effect was best using sym4 wavelet, the variance was 8.1177byte/s and 2.2316byte/s before and after signal de-noising, increasing 2.63 times, theoretical analysis and the actual matched, verifying conformity with the theoretical analysis.