一种新型信号奇异值预处理方法

A Novel Pretreatment Method for Elimination of Signal′s Singular Points

随着变频器等大功率电子开关器件的广泛应用,电信号中电磁干扰现象越来越严重,数据采集系统实时数据中的奇异点增多,且其分布特征不明显,对于低成本小系统来说,一些先进的数据处理方法无法实施,而传统均值滤波法和中值滤波法等滤波效果往往不理想。对此,提出了一种新型的信号奇异值的预处理方法,即PMF(预测均值滤波)法。该方法通过计算信号的前向均值和后向均值来分析信号的趋势,并将这个趋势作为判断奇异值的依据。如果是奇异值则将现有数据舍弃并根据前面判断的趋势对信号进行插值;如果不是则保留原数据,最后再对新的数据序列作均值滤波。应用结果表明,信号经过PMF法处理后其特征性能和平滑性得到大幅度提高,具有较好的实用价值。

Due to the popular application of power electronic switching parts such as invertors,the electric-magic noises in electric signals became more and more violent.The singular points in signals increased,but their distribution symposia are implicit.In small system with low cost,some advanced signal processing methods were difficult to implement.And traditional filters like mean filters or median filters are unable to get expected results.Therefore,this paper mentioned a novel pretreatment method named as PMF（Prediction Mean Filter） for elimination of signal＇s singular points.This method analyzed the trends of the signals firstly through calculating the forward and backward mean values of the signals.Then,based on the trends acquired above,make decision on whether the point being processed is singular point or not.If yes,replace the point with a new insert one.Otherwise,retain the point.Finally,reprocess the data with mean filter.The results show that the signals treated became more symbolic and slider.