The inspection of arrhythmia using the Holter ECG is done by automatic analysis. However, the accuracy of this analysis is not sufficient, and the results need to be correct by clinical technologists. During the process of picking up one heartbeat in an automatic analysis system, an incorrect detection, whereby a non-heartbeat is picked up as a heartbeat, may occur. In this research, we proposed the method to recognize this incorrect detection by use of a Support Vector Machine (SVM). When the learning results were evaluated on the ECG wave data from the one hundred subject’s heartbeats, this method correctly recognized a maximum of 93% as incorrect detections. These results should dramatically increase the work efficiency of clinical technologists.