基于多小波基多信源融合异常值剔除方法研究

Research of Eliminating Abnormal Datas in Multi-Source Fusion Based on Multi-Wavelet Basis

提出了一种基于多小波基多信源融合和偏度分析相结合的异常值剔除方法。基于多个小波基的数据融合算法,先对各信源的数据进行多个不同小波基的多尺度分解,对相同小波基分解的信号在多尺度上加权融合,之后进行不同小波基的逆变换得到重构信号,将重构信号融合出的结果作为目标状态估值,对各信源的异常值进行检测,将偏度分析与检测门限相结合,实现了对异常值的剔除。实验结果表明,该方法准确、高效地实现了对各信源异常值的检测与剔除,提高了数据处理的精度,且易于工程实现。

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.