Fractal Feature Based ECG Arrhythmia Classification

We propose a method for the classification of ECG arrhythmia using local fractal dimensions of ECG signal as the features to classify the arrhythmic beats. The heart beat waveforms were extracted within a fixed length window around the R-peak of the signal and local fractal dimension is calculated at each sample point of the ECG waveform. The method is based on matching these fractal dimension series of the test ECG waveform to that of the representative ECG waveforms of different types of arrhythmia, by calculating Euclidean distances or by calculating the correlation coefficients. The performance of the classifier was tested on independent MIT-BIH arrhythmia database. The achieved performance is represented in terms of the percentage of correct classification (found to be 99.49% on an average). The performance was found to be competitive to other published results. The current classification algorithm proved to be a computationally efficient and hence a potential technique for automatic recognition of arrhythmic beats in ECG monitors or Holter ECG recorders.