Rett syndrome is a disease that involves acute cognitive impairment and, consequently, a complex and varied symptomatology. This study evaluates the EEG signals of twenty-nine patients in order to implement an effective classification method to find the optimal artifact reduction strategy in each case. The classification has been made based on the mean and standard deviation (SD), allowing to differentiate patients with stereotyped and constant movements from those with a greater number of spasm or sudden movements. Since the various signal patterns may require diverse treatments, two artifact reduction methods have been analyzed. The first oneis based on the distribution, using again the mean and SD, and the second one is based on an energy function which, theoretically, should be more robust to outliers and more stable to signal to noise ratio. The results confirm the existence of three groups of signals differentiated by having: low mean and low SD, high mean and low SD and high mean and high SD. However, despite finding three different patterns, the energy-based method is the one that works best for all them, offering adequate adaptation to each type of signal without losing robustness and stability. In conclusion, its implementation improves the detection of outliers without compromising artifact-free data segments, which allows to maintain the quality and quantity of the records.