In this paper, the multifractal detrended fluctuation analysis method was used to study the multifractal characteristics of atrial premature beat (APB) signals, of premature ventricular contraction (PVC) signals and of normal ECG signals. By analyses of Hurst index, Renyi index and multifractal spectrum, three kinds of signals were noted to have different degrees of long-range correlation and multifractal characteristics. Normal ECG signals had the strongest fractality, the PVC beats had stronger fractality and the APB beats had the weakest fractality. When the fluctuations function order was positive, the three kinds of signals showed distinct long-range correlation properties. These findings are of good reference to diagnosing and distinguishing between PVC and APB signals in clinical medicine.