Extracting the features and methods of separating the letters and mental words through the EEG signals

Background and Aim : One of the problems that preoccupied the researchers today is how the perceptions are shaped and how to detect it from the brain. Perhaps one of the most important applications is the BCI for communicating the remote from the mind or even writing based on imagine. In order to speak first, person process their desired words in the mind and then speak. If a person has been unable to speak for any reason, how would it convey the desired word to those around you This is while the individual has processed the desired word in mind, but is unable to express it. So what we think is happening in the brain. So, if the brain s activities are detected through examining nervous signals while thinking, it may be possible to find a trace of what people are imagining.Methods : In this study, persian letters and words have been used. The words in the minds of the subject can have different categories. For a word, there may be some equivalent in the mind of the subject and there is no equivalent for another word. For this purpose, familiar and unfamiliar words as well as the letters used as a criteria for comparing two types of words were imagined and recorded during the two imagine and vision tests of brain signals. So that in the vision test, the subjects saw 11 words and letters and imagined them in other test.Results : After the signals were recorded and analysed in the two time and frequency domains, finally features were found to separate imagined letters and words. In the domain of time, when thinking to the word, the signal has a defined rhythm and volatility associated with the alpha and beta bands. On the other hand, the corresponding word signal diagrams also had phase differences so that unfamiliar words signal was recorded earlier. In the frequency domain, it was observed that by the beginning of the imagine, the Alpha power was increased that Its increased for unfamiliar word, familiar words, and letters, respectively. However, within the range of the stimulus presence, the Alpha power is reduced, suggesting a focus. As well as the β band ratio to the Alpha as a measure of focus, it was observed that this ratio is increased for letter, familiar and unfamiliar words, respectively.Conclusion : This study was based on the analysis of mental words, and in the first step between familiar words and unfamiliar words, it was found finally that the best way to separate the words and letters is the high beta range, which the beta to alpha ratio for letters is much lower than words. Also, the low beta range(12-20Hz) is the appropriate range for the separation of familiar and unfamiliar words. The accuracy of these data was obtained by t testing and p < 0.05, and it was observed that all these traits apply to the larger statistical community.

کلیدواژه ها:

Mental words ، Imagine ، Electroencephalography ، Classification ، Brain bands