The Wavelet Transform-based Ecg Processing And Analysis Of Research

Cardiovascular diseases are major diseases that endanger human's life. The application of Electrocardiogram (ECG) is essential for the clinical diagnosis of cardiovascular diseases. The use of computers for accurate and speed ECG diagnosis has been a subject fervently pursued by both internal and external researcher, and the automatic ECG classification is also a difficulty in the ECG analysis. In each kind of cardiovascular disease, Premature Ventricular Contraction (PVC) is the most familiar arrhythmia. Real-time, correct PVC detection is an important technology to improve detection of arrhythmia and the performance of monitoring unit, dynamic ECG analysis system.In this paper, according to characteristics of ECG signal and wavelet transform, on the basis of predecessors' achievement, the following aspects have been deeply studied:1. Pretreatment of ECG: a method called the wavelet threshold denoising is used to remove the interference of power line, electromyography and baseline drift, and a favorable effect is obtained through the simulation.2. Detection of QRS complex: according to the relationship of singular point, Lipschitz exponent and wavelet transform modulus maximum, a method called zero-crossings is introduced to detect QRS complex. With properly selected mother wavelet, 99.40% accuracy of QRS complex detection, which is certified by MIT-BIH arrhythmia database, is reached.3. Automatic classification of ECG: A neural network based on wavelet for ECG classification is established. By adjusting the scaling and shifting factors of wavelet function, which replaced the formal nonlinear fundament function, the network extract the characteristic of input signal automatically, then classify the it. After trained with a large number of the ECG samples, the network not only has a high accuracy (97.5%) to the training samples, but also has a good accuracy (94.7%) to the samples not trained before.