QRS波的多尺度融合检测算法

QRS Multi-Scale Fusion Detection Algorithm

提出一种基于模糊积分的小波变换多尺度融合检测算法。算法首先对心电信号进行 Haar二进小波分解 ,并在每个尺度上输出各自的判决向量。在各个尺度的检测中采用了软决策方式 ,使决策形式更加合理。小波变换后各个尺度上的信息存在互补性 ,采用模糊积分的融合方法可以充分发掘和利用各个尺度的信息。在小波分解后 1尺度和 2尺度的效果较差 ,因此只选择 3至 6尺度的决策作为模糊积分融合算法的输入。尺度的检测效果是确定模糊积分密度的依据 ,因此融合算法不仅考虑了各尺度的决策结果 ,还考虑了各尺度之间的差异。实验表明 ,模糊积分融合算法可以有效提高检测能力。

This paper presented a fuzzy integral based multi-scale fusion QRS detection algorithm which Haar wavelet transform is applied to analyze ECG signal. Every wavelet transform scale gives its decision independently. Soft decision is adopted on every scale, which makes the decision more reasonable. Because information on different wavelet transform scale is dissimilar and complementary, fuzzy integral is applied to utilize the information sufficiently. Scale one and two are neglected due to its poor performance, only decisions of scale three to six are chosen as input vectors of the fuzzy integral fusion algorithm. Fuzzy density is defined on performance of scale three to six, so the fusion algorithm not only considers the result of every scale, but also difference of detection ability among scales. Experiments performed under different noise levels showed that the algorithm could improve detection capability significantly.