In this article, firstly we present a new approach for detecting QRS complex which integrates the multi-scale idea of watershed into differentiation and template matching technologies. In order to make this algorithm accurate and robust, simultaneous multi-channel detection and fusion techniques are also adopted. Then, several features such as RR intervals and QRS durations of a certain ECG segment are extracted. Finally, we developed a branch-logic tree to automatically give interpretive statements for this ECG segment. The test data includes digitized paper-based clinical conventional ECG segments, MIT-BIH Arrhythmia Database, ECG from a Multiparameter Patient Simulator and some real data measured by our own Portable Intelligent ECG Monitor (PIEM) from several patients. The experiment results show that this novel method is efficient, robust and accurate in cardiopathy diagnoses.