Decoding Methods For EEG Signal Of States Of Brain Activities And Application Systems

Brain is one of the most complex organs in nature. And, it is the control center ofwhole body, involved in activities such as emotion, cognition, sense, and behavior. Inves-tigation in structure, function, and activity representation of brain will reveal mysteries ofbrain, and provide a solid basis for treatment of brain diseases. Therefore, research of brainmechanisms, in this dissertation, is of important significance and practical value. In thisdissertation, we investigated the expressions of cerebral activities under circumstances ofdifferent statuses of thoughts, and revealed the mechanisms of brain activities. According tothe mechanisms, we developed recognition algorithms to classify different states of activitiesunder circumstances of different statuses of thoughts. An important application for decod-ing brain activities is to build a brain computer interface. Brain computer interface (BCI)allows people to directly communicate with external world without the traditional pathwaywhich needs helps of external nerves and muscles. We systematically studied on the wholeprocess, including revelation of cerebral activities under circumstances of different statusesof thoughts, development of recognition algorithms for classification of brain activities, andbuilding of brain computer interface systems.The main contributions of this dissertation are given as follows:1. Existent studies in the field of BCI mainly emphasize on improving BCI system byprogress in BCI model, but ignoring human neuro-feedback adaptive adjustment whichis considered as the other aspect to improve performance of BCI. In our bilateral train-ing paradigm, we take human factors into consideration during the process of training.Subjects are requested to adjust their brain activities according to neuro-feedback re-flecting ongoing neural activities and to try their best to correctly achieve tasks ofmotor imagery. BCI model training and human training are implemented alternately.Human and BCI system are both trained in order to adapt to each other and graduallyattain a dynamic equilibrium. Hence, the reliability of system is improved due to themutual adaptation of human subject and BCI model. 2. A problem within the traditional training paradigm is that a trial is probably misla-beled if the participant don’t imagine the required type of motor imagery accordingto the given cue. Then, those mislabeled trials affect model training, which leads toa decrease of validity and performance of that model. Hence, we proposed an activetraining paradigm to solve this problem. With active training paradigm, the partic-ipants have a chance to confirm or disconfirm the label of the preceding trial whenthey feel it is necessary. Our paradigm makes participants more actively in the ex-periment than the traditional paradigm. Thus, participants are able to influence thetraining effect not only through modulation of their brain activities, but also throughactive engagement in labeling trials.3. In order to investigate effects on event-related potential (ERP) caused by different scalemotor imageries, we designed comparative experiment between index finger motorimagery (small scale) and arm motor imagery (large scale). Performances of differentscale motor imageries are evaluated in terms of classification accuracy and spectralpower. The aim of this experiment is to explore a mode, which can induce ERP moreeasily. We expect to provide a positive experience for researchers who engage in motorimagery experiments and make them more expediently select suitable type of motorimagery.4. A delayed movement paradigm was designed to investigate brain activities and mech-anisms in the human brain during planning of reaching movements. The ERP ampli-tude shows ipsilateral positivity and contralateral negativity in the PPC with respectto the movement direction. EEG spectra in the α band exhibit a distinct pattern ofipsilateral spectral augmentation and contralateral spectral depression with respect tothe movement direction at a later stage. During this later stage, direction-related ac-tivities (slow waves) in the medial frontal and medial parietal areas are found. Theresults of ERP and spectral power analysis showed that EEG signals generated in thePPC areas carry complementary information about intended movement direction in thebrain processes of visuomotor transmission and visuospatial attention. Hence, decod-ing accuracy could be improved by combination of ERP features and spectral powerfeatures.5. We designed and developed assistive wheelchair system, multi-person car racing sys-tem, web browsing system and EEG visualization system according to a variety ofapplication purposes. Assistive wheelchair system is used to help people with severe motor disabilities restore movements, without assistance of caregivers. multi-personcar racing system is a prototype of BCI-based game, and could become a new inputmethod in the future game development. In addition, a web browsing system is devel-oped to directly access network information from brain. In order to display essentialinformation contained in EEG signals, we developed EEG visualization system.6. We developed a training platform of motor function rehabilitation, which has beentested in hospital. This platform could help patients restore or rebuild the motor func-tion by motor imagery and external FES. Activation in motor cortex area through motorimagery and stimuli in peripheral nerves and muscles through FES result in restoringor renewing neuron connections, and rebuilding the pathway between brain and limb.In summary, this dissertation investigated the mechanisms in human brain under cir-cumstances of specified tasks, such as delayed reaching movements, motor imagery. Twotraining paradigms are proposed to improve the performance of BCI training. In addition, afew BCI application prototypes have been developed for a variety of application purposes.A rehabilitation platform is built for the use of motor function rehabilitation training. Thework in this dissertation provides brain mechanisms related to specified tasks and applica-tion prototypes, which would be positive in brain research and development of practical BCIsystems.