MEMS gyroscope has the advantages of small volume, low cost and easy integration, but its low accuracy greatly limits its application in practice. The measurement accuracy of MEMS gyroscope can be improved by using multisensor fusion technology for error compensation, so people have proposed many kinds of data fusion methods for improving the measurement accuracy of the MEMS gyroscope. In this paper, the multiscale fusion method, the Kalman filter fusion and the wavelet threshold fusion method, are compared and analyzed. Theory analysis and experiments results show that, comparing with the Kalman filter fusion and the wavelet threshold fusion method, the multiscale fusion algorithm has better performance on standard deviation, signal to noise ratio, power spectrum, and the Allan variance and so on, and it has a wider scope of the application.