The Feature Extraction Of ST Segments Based On Wavelets Transform And Its Application In Coronary Heart Disease Diagnosis

The dynamic electrocardiogram (ECG) is one of the most popular and basic diagnosis methods for coronary disease. It is more accurate and quicker than other diagnosis methods, and it has no hurt to people. With the development of computer technology, the automatic analysis of ECG is always the topic which scholars focus their research on. In Clinic application, ECG automatic analysis is still in the grope stage because its character extractive algorithms have low precise and its identification methods are unsatisfactory. Therefore, it is important to explore a more precise extractive algorithm and a more reliable identification method, so we design an extractive algorithm and identification methods of ST segments' shape in the paper. The accomplished work has three parts as following:(1) An algorithm is developed to extract the characters of ECG signals accurately. First, ECG signals are decomposed by aTrous Algorithm using dyadic spline wavelets, and the relation between the characteristic points of ECG signals and the modulus maximum pairs of the signals' wavelet transformation (WT) is illustrated. Then, taking advantage of the multiple resolution ability of WT, we develop a scheme to identify the R-wave and ST segment's fiducial points at different wavelet scales automatically. The proposed method is demonstrated by the datum from the standard MIT/BIH ECG database and the data generator. It is shown that the proposed algorithm has the virtues of strong anti-jamming capability and high precise.(2) An identification method is designed to classify the shapes of ST segments. Based on the accurate extraction of ECG signal characters, slope law united with function fitting method is applied into the classification of ST segments shapes. The experimental results show that the method can identify the six shapes of ST segments correctly. This method is simple and quick. However, it is influenced by the extractive accuracy of characters and the signal noise greatly.