Abstract: It is difficult to identify the ideal signals in the self-adapting digital filter. In order to solve this problem, taking account of the mean squared error and the characters of high-frequency signal, this paper puts forward a new self-adapting digital filter. 252 real data are used to test the effects of the self-adapting filter. In order to evaluate the filter effects, Back-Propagation（BP） neural network and Support Vector Machine（SVM） are respectively employed to classify the results of the data sets. The experimental results show that this method has the best performance among all filtering methods.

Key words: self-adapting filter, mean squared error, high frequency, compromise evaluation

摘要： 自适应数字滤波中理想信号通常难于确定。针对这一问题，根据均方误差和高频信号的特征，将二者结合起来考虑，提出一种新的自适应数字滤波的方法。应用该方法在252组真实实验数据中进行相应的自适应滤波测试，并对滤波结果分别采用BP神经网络和支持向量机两种分类方法进行分类测试。实验结果表明，新方法具备良好的滤波效果。

关键词: 自适应滤波, 均方误差, 高频, 折中评价