基于冗余离散小波变换的信号配准及分类

摘要

离散小波变换(Discrete wavelet transform,DWT)对输入信号的平移敏感,当输入信号间存在平移时,基于DWT的信号分类会受到负面影响.本文提出一种基于冗余离散小波变换(Redundant DWT,RDWT)的信号配准及分类方法,克服了DWT的平移敏感性,解决输入信号间存在平移变化的信号分类问题,同时实现了信号配准以及平移不变小波系数提取.对基准信号作离散小波变换.对测试信号作冗余离散小波变换,得到测试信号的平移所对应的DWT分解,计算其同基准信号DWT分解之间的相似性,根据相似性大小对该测试信号分类并确定其相对基准信号的平移量.利用ECG(Electrocardiograph)信号的仿真实验证明了本文所提方法的有效性.%Discrete wavelet transform (DWT) is sensitive to the translation/shift of input signals, so its effectiveness could be negatively impacted when it encounters translation among signals. This paper proposes a redundant DWT (RDWT) based signal registration and classification method that can solve DWT's translation sensitivity, fulfill signal classification when there are translation variances among signals, and simultaneously accomplish signal registration and translation invariant wavelet coefficient extraction. To the reference signal, we perform DWT on it. To the test signal, we perform RDWT, obtain the DWT result for its translation, and compute the similarity between the translated test signal and reference signals based on the DWT results. Then, we classify the test signal and determine its translation with respect to the reference signal according to the similarity criterion. Simulation results on the ECG signals prove the effectiveness of our RDWT based method.