基于心搏聚类的Holter运动伪差段快速识别算法

为了能够在Holter分析系统中快速地标识出夹杂着运动伪差的时间段,提出一种基于心搏聚类的运动伪差段识别算法.先对心搏的QRS复合波面积进行直方图统计,得到心搏特征的主导分布；结合心搏形态,采用基于“轮廓限围”和“相关差值”的模板匹配算法完成心搏聚类；根据聚类结果得到的心搏凌乱特征曲线,自动识别出存在运动干扰的时间段.从医院采集了运动伪差比较明显的10例24 h Holter临床数据进行算法测试,实验表明：24 h的心电数据平均耗时不到1 s,识别结果与人工标注基本吻合.

Beats clustering based algorithm for fast recognition of motion artifact sections in Holter system

An algorithm based on beats clustering was presented to fast recognize the electrocardiography(ECG) signal sections that include motion artifacts in Holter system.Firstly,the dominant distribution of beats features was obtained by analyzing the histogram of QRS area of all beats.Then,a template matching technique based on "contour limit" and "relevant difference" was used to cluster beats with similar shape.Finally,the motion artifact sections were automatically identified according to the beat clutter characteristic curve obtained from the clustering result.The algorithm performance was evaluated on ten 24 h ECG clinical cases with obvious motion artifact that collected from the hospital.The results showed that the average analysis time of 24 h ECG data was less than 1 s,and the recognition results coincidented with the manual annotation.