Abstract: Based on multi-scale data analysis in wavelet domain,different types of noises for MEMS(mirco eletro mechanical system) gyroscopes are considered.Through the noise and analysis based on wavelet multi-scale entropy,the appropriate wavelet is selected for data fusion.The experiment study is carried out from the statistical point,and the superiority of this approach is proved mainly by wavelet entropy.At the same time,the correctness of multi-scale data fusion of the optimal wavelet basis is verified with the probability distribution,and the technology problems in engineering for the multi-scale sensor data fusion approach are solved.