Application of wavelet transform in extracting ECG feature

It is an essential and vital function for electrocardiogram (ECG) analysis instrument to exactly extract the information of QRS complexes. In this article a dyadic spline wavelet is taken as mother wavelet, and a multi-scale wavelet transform is applied to the ECG signals. And according to the characteristics of transformed signal it is found that the signal at 3(rd) ( S = 2(3)) scale is only for extracting the QRS characteristic waves of ECG signals. There exists the corresponding relationship between a characteristic point (R peak point) of the ECG signals and a zero-crossing point of the modulus, maximum pair. Combining. with some compensation tactics (e.g. sectional threshold values and slope criterion for the modulus maximum pair) we present an algorithm for accuracy extracting QRS complex. The algorithm is applied in experiments to massive data from MIT/BIH arrhythmia database and clinical, and the results indicate that even for the data contaminated by serious noise it is easy to extract the feature information of QRS complexes. The locating precision (no more than one sampling points error) and analyzing accuracy (no accumulative error) are high, especially for R peaks.