In the process of using the wavelet domain multi-scale data fusion algorithm to fuse MEMS gyroscope signals, because the gyroscope signal contains a trend term, the dynamic data of the gyroscope, especially when it contains jump data, causes the fusion value to be delayed compared to the original data of the gyroscope. In order to solve these problems, a new MEMS gyroscope data fusion method switching scheme is proposed, first using the sliding average method to extract the trend term of the gyroscope signal, and then performing the fusion processing on the gyroscope signal after removing the trend, using the cumulative and control chart algorithm and the preset threshold is used to determine whether the gyro signal is transitioned as a switching condition of the fusion method. If there is a transition, the orthogonal basis neural network data fusion algorithm is used to perform fusion processing on the collected gyro data. The experimental results show that after using this scheme, the problems caused by the original single-scale multi-scale data fusion in the wavelet domain are solved, the variance is reduced to 0.00351 in the constant rate experiment, and the effect is also significant in the jump experiment, the purpose of improving the overall output accuracy of the MEMS gyroscope during the measurement process is proved, and the effectiveness of the scheme is proved.
|