Abstarct: ALS is a kind of disease that damages the neural system in the body and causes disability in speaking and normal movement. In a severe situation it might even affects the control of eye movement which is called complete locked-in syndrome. Fortunately, this disease does not affect the brain. Therefore, the only way to help these people is developing a system so that the patient can communicate to the environment through the brain. Brain computer interface is one such well-known systems. One of the most important methods used in BCI systems is taking advantages of ERP components in which is produced as a result of brain response to visual, cognitive or sensory events. P300 which is one type of ERP, has been used in several interfaces such as P300Speller baxse on overt attention, and GeoSpell baxsed on covert attention. These interfaces requires not much training and therefore are very effective in improving life quality for disabled people. In this thesis, several methods baxsed on frequency selective filters, independent component analysis, and common average referencing have been used to improve the P300 event related potentials. These methods have been used either individually or as a combination to other techniques under P300Speller or GeoSpell protocols. After data processing using the above methods, simultaneous synchronized averaging is used for extracting P300. Then, statistical assessments for comparison of the obtained results. The results reveal that even without structural difference between these two paradigms, there is no significant difference, and GeoSpell which is specifically designed for ALS can effectively produce P300. Also after detecting the P300 components and comparing it with the reference paper, it was observed that in GeoSpell interface, the Latency and Peak of P300 signal were 378.23 ms and 22.007 μV, respectively, which was better than the reference article with Latency of 453.68 ms and the Peak of about 1.8 μV. In the P300Speller interface, the Latency and Peak of P300 were 346.95 ms and 25.93 μV, respectively, which was better than the reference article with Latency of 366.76 ms and the Peak of about 2.5 μV, too.