Fractal Random Walk and Classification of ECG Signal

This paper presents a new nonlinear method to analyze ventricular arrhythmia(VA) and sinus rhythms(SR). The problem is introduced from the discussion of Fractal Random Walk characteristics of ECG signal. Further, the fractal analysis is used to distinguish ventricular flutter(VFL), ventricular fibrillation(VF), ventricular tachycardia(VT)) and sinus rhythms(SR) from the raw electrocardiogram(ECG) data. The method has a three step processing. First, calculating the slope of permutation entropy(PE) to detect the onset of ventricular arrhythmia; Second, using regularization dimension(RD) to classify SR, VFL and VT/VF; Finally, according to multifractal spectrum(MS) area to distinguish VT and VF. Four databases are used to detect the method, and the accuracy of every step is 93.33%, 100% and 98%. As a whole, the accuracy of detecting onset of ventricular arrhythmia and confirming which ventricular arrhythmia is, is VFL 93.33%, VT and VF 91.47%.