Time irreversibility of electric cardiac graph(ECG) signal can reflect the physiological function and the health condition of the heart. In this paper,we propos a symbolic relative entropy method (first symbolizing and then determining the time irreversibility) to analyze the three kinds of ECG signals, i.e., normal sinus rhythm, ventricular fibrillation and sudden cardiac death from the MIT-BIH database. The results show that three kinds of signals have different time irreversibilityies, and the time irreversibility range is largest for the normal sinus rhythm, the second largest for the ventricular fibrillation, and smallest for the sudden cardiac death. Variance test results indicate that the irreversibilities of the three signals are significantly different. The results show that the relative entropy method can play a positive role in heart disease detection and diagnosis.