There has been great interest with the Motor Imagery (MI)-based brain-computer interface (BCI), recently. We developed the Novel Adaptive threshold for feature extraction. A string grammar fuzzy K-nearest neighbor is developed by incorporating 2 types of membership value into string grammar's K-nearest neighbor. The new algorithm used for the disabled or the patients who are physically unable to move with used to distinguish the order. In this experiment we check the decision to lift the arm. We apply these two-string grammar fuzzy K-nearest neighbors in the brain classification system. The system provides 88.14% in our dataset.