Anomalous intervals in RR sequences of electrocardiogram (ECG) introduce practical problems in heart rate variability (HRV) analysis, trauma evaluation, and other clinical applications where ECG artifact rejection is important. Impulse rejection filters (IRFs) have been used to identify anomalous intervals. However, traditional IRFs do not consider the influence of non-stationary time series of RR intervals, and cannot effectively identify anomalous intervals caused by ECG morphology changes. This study therefore improves the traditional IRF method, develops an ECG morphology feature-based template matching method, and develops a combination method of improved IRF and template matching for better identification of anomalous intervals in RR sequences. Four methods (IRF, improved IRF, template matching, and the combination method of improved IRF and template matching) are applied to the MIT-BIH arrhythmia database. 43 RR sequences that contain 30368 normal intervals and 2760 anomalous intervals are analyzed. Sensitivity and specificity analyses are performed to quantify identification performance. The sensitivity and specificity values are 86.1% and 87.0% for IRF, 92.3% and 94.9% for improved IRF, 90.7% and 95.5% for template matching, and 98.5% and 99.2% for the combination method, respectively. The results verify that the combination method is suitable for the identification of anomalous RR intervals.