Brain Computer Interface provides a new communication channel for people who have severe brain injuries. Among different types of BCIs, SSVEP-based one has been focused in recent years. In this type of BCI, selection of twinkling frequency of external visual stimulant and the distance between stimulants (in case of more than one stimulant) is so important. In this work, a SSVEP-based BCI with two external stimulants was designed. In order to determine the best twinkling frequency of stimulants and the best distance between them, the classification accuracy for seven different twinkling frequency pairs and five different stimulants distances was calculated. Two methods for feature extraction step were proposed and the Max classifier was used for classification in order to speed up the computational burden of classification step. Features were extracted from four different segment lengths (sweeps). The results showed that nearly in all sweeps and all inter-sources distances the frequency pair of 10–15 Hz has the highest classification accuracy among other frequency pairs, which is 92% in the inter-sources distance of 24 cm and sweep length of 3 seconds. In addition, the results demonstrated that the method 2 feature extraction technique outperforms the method 1. In addition, for determining the best sweep length, Information Transfer Rate (ITR) was computed and the results indicated that the sweep length of 0.5 second has the highest ITR, so would be practical in real-time applications of BCI.