ECG events detection and classification using wavelet and neural networks

An integrated system for ECG diagnosis that combines the wavelet transform (WT) for feature extraction and artificial neural network (ANN) models for the classification is proposed. By using the dyadic wavelet transform, the limitations of other methods in detecting ECG features such as QRS complex, the onsets and offsets of P and T waves are overcame. The ECG baseline is approximated using discrete least squares approximation. On classification, two paradigms of learning, supervised and unsupervised, for training the ANN modes are investigated. The backpropagation algorithm and the Kohonen's self-organizing feature map algorithm were used for supervised and unsupervised learning, respectively. The system is evaluated using the MITBIH database. The result indicates that the accuracy of diagnosing cardiac disease is above 97.77%. ECG signals can be classified, even with noise and baseline drift.