Electrocardiogram (ECG) beat classification is essential for timely diagnosis of serious heart disorders. Premature ventricular contraction (PVC) and atrial premature contraction (APC) are serious threats against human lives. In order to cope with difficulties in diagnosis, a mode entropy based technique is herein proposed to classify normal sinus rhythm (NSR), APC and PVC, As is well known that the accuracy of most methods for detecting NSR, APC, PVC, etc is now between 90% to 98%, but most of them need large data sets. After calculation of the mode entropy value of NSR, PVC and APC, the results were analyzed and compared. The detection method based on the mode entropy technique is proven to be rational and accurate, and it does not need large data sets.