Brain-computer Interface Of Website Browsing System Based On EEG Signals

Electroencephalogram (EEG) is spontaneity, rhythmic electrical activities from cells, whichcan reflect the consciousness of human effectively. EEG can be recorded by electrodes. We caninterpret brain signals and understand people’s awareness activities by analyzing EEG signals.Brain-computer interfaces (BCIs) collect EEG signals as input, which can achieve directcommunication and control between human brain and outside (a computer or external device),rather than relying on normal channels such as peripheral nerve and muscle tissue. BCIs providea new way for human to communicate with outside. Because of its important theoretical researchvalue and broad application prospects, BCIs cause attention of the world and become one of thehot research topics.The thesis delved into the EEG signals processing and established a set of online webbrowsing brain-computer interface system based on off-line analysis. Off-line analysis delvedinto the EEG acquisition and signal processing based on three EEG signals: steady-state visualevoked potential (SSVEP), α waves and motor imagery. Preparation and collection of EEG,electrodes position and the way of EEG channel link were illustrated in the signal acquisitionpart. In addition, designing experiment process and collecting three kinds of EEG signals werecompleted in this part. Original EEG often contains strong noise interferences. To get pure EEGsignals, signal preprocessing was essential. In the preprocessing part, FIR band-pass filter wasadopted to filter the original EEG signals. Besides, independent component analysis (ICA) wasadopted to remove ocular artifact mixed in motor imagery. EEG in the BCI system has higherrequirements for the signal processing algorithm because of its strong background noise, highlynonlinear and nonstationary. Many research institutions are working on the signal processingalgorithms in order to improve the control precision and speed of the BCI system. Featureextraction and pattern recognition is the key to determine the performance of the BCI system. Inthe feature extraction part, different feature extraction methods were chosen according to thecharacteristics and complexities of different EEG signals. Namely, fast Fourier transform wasused for feature extraction for steady-state visual evoked potential, while short-time Fouriertransform and wavelet packet decomposition were used for feature extraction for four-classmotor imagery. Then calculate corresponding frequency energy as the classification of featurevectors. In the pattern recognition part, linear classifier and support vector machine (SVM) wereadopted to classify the EEG signals after preprocessing and feature extraction. Select the idealfeature extraction and classification method by comparing classification results to provide basisfor online website browsing system. Finally, establish an online website browsing system on LabVIEW. Two experiment paradigms were designed with evoked EEG and spontaneous EEG.Through this online BCI system, participants can complete cursor control and simple webbrowsing tasks under two paradigms.