Electrocardiogram signal classification based on fractal features

Atrial fibrillation ECG signals have been classified with fractal features only. The fractal features -fractal dimension, mass dimension and lacunarities were estimated by a new box counting algorithm; called the true box counting method. The classification result and stepwise discriminant analysis for these fractal features were determined. It was seen that lacunarities based on higher mass moments were more important than fractal dimension and mass dimension. The results suggest further investigation of lacunarity features.