Comparison of the neural-network-based adaptive filtering andwavelet transform for R, T and P waves detection

摘要

An electrocardiogram is quite an important tool to find out moreinformation about the heart. Despite the presence of serious noise,reliable detection of the QRS complex, and T and P waves is essentialfor an exact ECG analysis system. ECG analysis methods can be dividedinto three functional groups: direct methods, transformation methods,and parameter estimation methods. For real-time processing nonsyntactic,neural network based detection is ideal. Both of the selected methodsaccomplish filtering of the ECG signal. Neural network based adaptivematched filtering is capable of learning and becoming time-varying.These filters estimate the deterministic signal and remove uncorrelatednoise with the deterministic signal. This method can produce betterresults than nonparametric algorithms. These advantages result in lowernoise sensitivity and required sampling rate and accuracy at the samedetection rate for all the ECG characteristic points