一种多阈值融合心音递归图的设计与应用

Design and application of a multi-threshold fusion recurrence plot of heart sound

为了有针对性地突出心音的某些特征,将心音信号用递归图的方式进行表征和识别。该研究讨论了心音与递归图的关系,给出适用于心音的递归图获取方法;提出一种自适应阈值计算方法,重点分析如何利用这种自适应阈值构建一种多阈值融合心音递归图,使这种心音递归图能最佳的表现心音特征,并从多阈值融合心音递归图中提取D^′2/S^′2值和灰度共生矩的4种纹理特征:能量、对比度、相关性和熵;利用支持向量机进行心音分类识别。实验结果表明,对研究中使用的两个数据库,不用进行S1,S2分段预处理,平均识别率可达到90%以上,并将这个方法成功应用于心血管健康评估系统。

To show some characteristics of heart sound,the feature of heart sound signal was characterized by a recurrence plot.Firstly,the relationship between heart sound and a recurrence plot was discussed,then a recurrence plot acquisition method for heart sound was given.Secondly,an adaptive threshold calculation method was proposed,then a multi-threshold fusion recurrence plot of heart sound was constructed,so that the recurrence plot of heart sound could best present the characteristics of heart sound.Next,four kinds of texture features of gray-level co-occurrence-matrix and the value of D^′2/S^′2 were extracted from the multi-threshold fusion recurrence plot of heart sound.Finally support vector machine was used to realize identification.The experimental results show that the average accuracy could reach to 95%without performing of S1,S2 segmentation.The method had been successfully applied to a cardiovascular health assessment system.