We designed software system to detect intermittent arrhythmia from the patients automatically. To overcome the difficulties on real-time processing, we used four variavles; heart rate, QS width, morpholigy and axis. Templete generation routine was designed to produce patient's individual normal ECG pattern and store in the computer. The most important point of automatic detection is to discriminate the Q and J point. It is done by linear phase differentiation, the point that five point gradient is within a minimum value of 6Hz curve gradient is defined as Q and J points. The morphology calculation is simplified by simple addition and substraction, making real-time processing by modified cross correlation method. We find that both value of morphology and QS width will be used as a parameter of datermining the abnormalities of ECG signal automatically.