Wavelet Analysis Based Signal Denoising Research In MEMS Gyroscope

The drift error is the main error source in Gyro inertial navigation system, and compensating gyro drift error effectively is the key point to ensure the accuracy of inertial navigation system. MEMS gyroscope drift error is random non-stationary, weak linear and slow time-varying, which is difficult to be compensated in the inertial navigation system with simple methods for the influence of uncertain external environment factors. Wavelet analysis is particularly suitable for non-stationary signal processing for its multi resolution attribute. In this paper, wavelet based analysis is applied to eliminate the MEMS gyro signal noise base on attributes of MEMS gyro signal. The wavelet transform modulus maxima denoising method, inter scale correlations of wavelet transform denoising method and wavelet threshold method are analyzed, and compared the effects of the three methods in practical applications. The main contents include:Firstly, background theories and application of wavelet thansform methods for MEMS gyro signal denoising are analyzed. The basic theories of continuous wavelet transform, discrete wavelet thansform, dyadic wavelet transform and wavelet multiresolution analyses are expounded. Especially, the principles, algorithms and characteristics of wavelet transform modulus maxima denoising method, inter scale correlations of wavelet transform denoising method and wavelet threshold method are emphatically introduced.Secondly, based on the MEMS gyro simulation signals which are generated by MATLAB, wavelet transform modulus maxima denoising method, inter scale correlations of wavelet transform denoising method and wavelet threshold method are applied to eliminate signal noises. Optimal wavelet basis and scale are concluded by the three methods, and Birge-Massart threshold method with scale 12 is proved the best for denoising, which eliminats most of the signal noise, keeps 94.91% signal energy, reduces standard deviation from 1.3270(°/s) to 0.0477(°/s) and promotes the signal stability.Thirdly, MEMS gyro experiments are carried based on test system, and angular velocity data of statistic, uniform rotation, and variable rotation are sampled. With the parameters that concluded in MATLAB simulations above, the sampled data is denoised with wavelet analysis methods. The results indicate that Birge-Massart threshold method shows the best denoising effect, which almost eliminats all the noise and keeps the highest energy ratio. At the same time, Birge-Massart threshold method reduces standard deviation from 1.3270(°/s) to 0.0477(°/s) for the gyro static signal; reduces standard deviation from 0.2169(°/s) to 0.0138(°/s) for the gyro uniform rotation signal; and almost eliminats all the noise for the ununiform rotation signal. In this paper, simulation signals generated by MATLAB and experimental signals sampled by MEMS gyro test system are denoised with different wavelet analysis methods, and it is proved that wavelet analysis based denoising methods have meaningful application value for promoting the performance of MEMS gyro.