动态心电图自动检测及分析方法的研究

【作者】 张文琼；

【导师】 刘肖琳；

【作者基本信息】 国防科学技术大学 ， 模式识别与智能系统， 2004， 硕士

【摘要】 论文以动态心电图自动检测及分析为背景,研究了基于小波变换的自适应阈值消噪方法,通过分析比较四种白适应阈值选取规则,用基于Donoho固定阈值的小波去噪法滤除了心电图工频及肌电干扰,取得了很好的滤波效果。 针对基线漂移噪声问题,本文提出了一种利用小波变换逼近信号进行滤波的新方法。通过对ECG信号进行多尺度分解,利用分解后所得的逼近信号充分逼近心电图中的基线漂移噪声的特性,滤除心电图中的基线漂移分量。实验证明该方法能在保证心电图特征无损害的情况下充分地滤除心电图基线漂移噪声。 在心电图特征检测过程中,针对QRS波群检测问题,改进了原有的小波检测方法,结合原始信号和小波系数两方面的信息加以检测,达到了99.99%的检测正确率。对于P、T波的检测问题,采用在特定区间内通过斜率和幅值加以检测的方法,对正常心电图检测效果较好,但不适用于不规则的和病态心电图。 最后,设计了基于决策树的心律失常自动分析方法,并以窦性心律和异位心律分类决策树的构建为例,说明了心律失常决策树的构建方法及自动分析规则的产生过程。更多还原

【Abstract】 Under the background of the automatic analyses system of the dynamic electrocardiogram, a method of de-noising via wavelet threshold was studied. By comparing four threshold-select rules, the Donoho immovable threshold was used to eliminate the 50Hz interference and muscle artifact of the electrocardiograms, which leads to excellent outcomes.A new method using wavelet approximation to remove the electrocardiogram baseline wander is proposed in this dissertation. Via the multi-resolution analysis of the original signal, the baseline wander of electrocardiogram can be characterized almost fully by the approximation signal. Then the baseline wander can be removed by dealing with the approximation signal. Applying this method to the real 30min-ECG-logs provided by MIT-BIH Arrhythmia Database, the experiment results show that it can effectively remove the electrocardiogram’s baseline wander, and have little harm to the other elements of the signals.During the process of character detection of the electrocardiogram, a method based on wavelet was improved to detect the QRS complex. It makes use of not only the wavelet coefficients, but also the original signals to find the accurate position of the peak wave of QRS complex. In the experiments, 99.99% signals can be detected exactly by using this method. After QRS detection, the P wave and T wave is detected by the information of the amplitude and slope in certain section. This method works well for normal signals yet it isn’t fit for the unmoral signals and diseased signals.Finally, an automatic analysing method of Arrhythmia based on decision tree was designed in this dissertation. The constructing of the decision tree and the producing of diagnostic rules were shown by the example of sinus rhythm and ectopic rhythm decision tree.更多还原