Abstract: Concerning the low accuracy and easily being interfered by noise of ECG diagnosis of ST-segment, a ST-segment detection algorithm based on least-square polynomial fitting was proposed. Firstly QRS complex was detected by dyadic spline wavelet through Mallat algorithm, then the characteristic points such as T wave, onset of QRS complex, J point, onset of T wave were detected. These characteristic points were used to judge the direction of ST-segment, and then the ST-segment was classified into line type and curve type, finally the slope of line type ST segment and concavo-convex direction of curve type ST-segment were determined through using polynomial fitting algorithm. This algorithm was certified by the simulation experiment on the signals of MIT-BIH database, and the accuracy of ST-segment shape recognition was more than 90%. The experimental results show that the algorithm reduces the noise interference of characteristic detection of ST-segment, and improves the accuracy of ST-segment detection.

Key words: QRS complex, ST-segment, wavelet, characteristic parameter extraction, least-square polynomial fitting

摘要： 针对目前心电图ST段诊断准确度不高,容易受到噪声干扰的情况,提出了一种基于最小二乘多项式拟合的ST段形态识别算法。首先利用二次样条小波经过Mallat算法检测出心电信号中的QRS波群,然后检测出T波、QRS波群起点、J点、T波起点等特征点,依此判断ST段偏移方向,并将ST段分为直线型和曲线形,最后通过多项式拟合算法来确定直线型ST段的斜率和曲线型ST段的凹凸方向。通过MIT-BIH心电数据库的数据文件的仿真实验验证了该算法用于ST段形态识别的准确度在90%以上,实验表明,该算法减少了ST段特征点检测过程中噪声的干扰,提高了ST段形态识别的准确度。

关键词: QRS波群, ST段, 小波, 特征参数提取, 最小二乘多项式拟合