Abstract: The method of eliminating abnormal datas in multi-source fusion based on multi-wavelet basis and skewness analysis is presented.First,The multi-wavelet basis and scale wavelet decomposition to the sensor signals is carring out. Then, the weighted dada fusion algorithm is implemented to signals of the same waveletbasis decomposition on multi-scale.Last,the signal is reconstructed by inverse transformation of different waveletbasis. The final fusion is the estimate of target state,moreover by combining skewness analysis and threshold detection,we can detect and eliminate abnormal datas．Experimental shows that abnormal datas can be detected and eliminated efficiently and quickly by this method,then data processing precision can be risen,it is also easy to realize in project.

Key words: Multi-wavelet basis, Data fusion, Abnormal datas, Elimination, Skewness analysis