Arrhythmia Analysis and Interpretation of ECG Signals

The various classification algorithms can be used to classify extracted features from ECG signal. High performance of classification depends on how well the vectors of features can be separated in the feature space. Proposed architecture presents ECG based arrhythmia classification with more robust features and Regression based classifier. It proposes an effective automated classification of cardiac arrhythmia using MIT-BIH arrhythmia database and Local Clinical dataset. Proposed Method have trained the Incremental Support Vector Regression Classifier with 320 samples of different arrhythmias. Proposed Method has been tested and compared with the most common classifier such as Artificial Neural network, Support Vector Machine and Minimum Distance Classifier. From The confusion matrix it is clear that our proposed algorithm works well for multiple class recognition problem. Proposed Architecture uses both time and frequency domain features for classification purpose. Due to use of higher order statistic our classification problem becomes simpler than traditional morphological feature. Proposed algorithm delivered high performance even with smaller learning data.