Studies And Design On SSVEP-Based Brain-Computer Interface

Brain Computer Interface (BCI) is a new method of human-computer interaction (HCI). It is used widely in many application fields with high practical value. This thesis aims to extract the signal features efficiently and swiftly, through the research of stimulation modes and signal processing methods. What's more, we improve the SSVEP performance on CRT monitor with the method of stimulating selections increasing.On the basis of a large number of literature review, the writer chooses SSVEP as the research subject, because of its fast processing speed and high feasibility. This thesis mainly involves two aspects. One is for the signal processing method, the other is related to experiment design.For the experiment design, firstly, because of the stimulation features of CRT monitor (square -wave stimulation), to avoid the interference of the square wave, we only choose the frequencies whose interval are integral multiples to the basis monitor flicker as the stimulation selections. Then analysis the experiment results of single and multiple selection frequency stimulation.Then, to the shortcoming of few stimulation selections in the CRT SSVEP experiment, this thesis brings up a concept of mixed frequency stimulation. The main idea of mixed frequency is to stimulate the testees with more than one frequency. When the frequencies are displayed one by one, we can find an interesting phenomenon, that is, when the difference of the two frequencies is more than 3Hz, we can extract those two frequencies precisely; while when the difference is less than 3Hz, the spectrum features appear at the average of the two frequencies and the sum of those two.For the signal processing (spectrum detection) research, this thesis tries to extract the spectrum features in two ways. One method is to combine the method of moving average filter and soft threshold method. This method is suitable to most of our testees except those whose brain wave is mainly alpha wave during the eye open state.To improve the method mentioned above, this thesis tries to combine the method of wavelet analysis and AR parameter model. From the analysis result, we can find that this method has more efficiency and accuracy than the one mentioned above,In general, this paper extracts the features of SSVEP effectively through the combination of wavelet and AR model, then increases the stimulations of CRT with the method of multi-frequency. From the extraction result we find that the multi-frequency has strong operability.At last, this thesis applies the researching result of multi-frequency stimulation to the visual dialing system. It can make the testees choose what they are looking at without any help of body movement. The multi-frequency method does increase the stimulating selection effectively.