中文摘要：

采用经验模态分解（EMD）、小波阈值去噪、神经网络自适应滤波方法对4种储粮害虫活动声信号进行滤波去噪,得到5组对比实验.经对比分析每种方法去噪后的效果,发现EMD方法提升的信噪比和均方根误差均比较稳定,神经网络自适应滤波方法去噪后信号有失真现象,且不稳定.结果表明：EMD方法较适用于储粮害虫活动声信号去噪,能够较好地消除含噪信号中的噪声,对其他储粮害虫的声检测同样具有应用价值.

英文摘要：

Four kinds of active acoustic signals of stored grain pests were de-noised using EMD,wavelet threshold de-noising and neural network adaptive filter.Five contrastive results under three different noise levels were gained.With analyzing and contrasting de-noising results of each method,the increased values of Signal to Noise Ratio and Root Mean Square Error of EMD method were relatively steady.De-noised signals of neural network adaptive filter method appeared anamorphic and unsteady phenomena.For the selected three wavelet base functions,the de-noising results of wavelet threshold de-noising method were inferior to the former two methods.The results show that EMD method is applicative in de-noising for active acoustic signals of stored grain pests and it can eliminate noise for noisy signal well.This result has the same applicative value to acoustic detection of other stored grain pests.