Abstract: Recent researches have proved that some types of J-wave syndrome are high-risk indicators of malignant arrhythmia, sudden cardiac death, and other illnesses. But the judgment of whether J-wave exists in most of the present researches relies only on electrocardiograms and clinical experience, which not only are time-consuming but can hardly avoid misdiagnoses. This paper, from the perspective of signal processing, proposes an approach for J wave auto-detection based on the variable step size support vector machine. After the optimal value of parameter [σ]of kernel function is fixed, SVM modelling is completed. Experimental results show the effectiveness of J-wave detection by this method, with the accuracy rate reaching high at 96.1%.

Key words: J wave, Support Vector Machine（SVM）, kernel function, feature vectors, variable step size

摘要： 近期研究表明J波综合征的某些类型成为恶性心律失常、心脏性猝死等疾病的高危预警新指标。但目前大多数的研究仅通过心电图纸和临床经验来判断是否存在J波，不仅耗时较长，而且难免会存在误判。从信号处理的角度出发，提出了一种基于变步长思想的支持向量机（Support Vector Machine，SVM）方法来实现J波的自动检测。利用变步长思想寻找最优的核函数参数[σ]并完支持向量机的建模。实验结果证明，所提的方法可以有效地检测出J波，准确率为96.1%。

关键词: J波, 支持向量机（SVM）, 核函数, 特征向量, 变步长