Font Size: a A A

Study And Implementation Of Online Motor Imagery Based Brain-computer Interface System

Posted on:2013-11-25Degree:MasterType:Thesis
Country:ChinaCandidate:X LiFull Text:PDF
GTID:2248330392958480Subject:Biomedical engineering
Abstract/Summary:PDF Full Text Request
Brain-computer interface (BCI) is a kind of system which enables brain tocommunicate with external environment independent of peripheral nerves and muscularpathways. BCI systems collect signals from the brain, classify through signal processingalgorithms and translate them into control commands for external devices. Motorimagery based BCI achieves its classification and control by extracting EEG featuresduring different limb motor imageries.As a subjective mental task, motor imagery does not require any outsidestimulation. It has wider range of applications, and is important for rehabilitationtraining of motor dysfunction patients such as stroke patients. Existing motor imagerybased brain-computer interfaces still have room for improvement in classificationperformance and task design. Seeking to further improvement of them is a meaningfulwork.In this thesis, based on the neurophysiological basis and the target feature of motorimagery, several commonly used and latest proposed methods for motor imageryfeature extraction and classification are introduced. Their performance is tested andcompared offline through actual motor imagery EEG data sets. Finally, iterativecommon spatial-temporal patterns (ICSTP) combined with support vector machine(SVM) is chosen as feature extraction and classification method for online motorimagery based brain-computer interface.Based on the offline research, an online motor imagery based brain-computerinterface system is built up. In this system, subjects control the horizontal movement ofa virtual ball by imagining right hand or left hand movements. In order to improve thesystem performance, it carried out asynchronous improvements on the task. New taskand visual feedback is introduced. The effectiveness of the system is proved throughonline experiments with subjects.One of the inherent shortcomings in motor imagery based brain-computer interfaceis a few identifiable tasks. To solve this problem, a new paradigm of motion-onsetvisual evoked potential is introduced to combine with motor imagery, forming ahybrid-controlled character input system which does not depend on strong visual stimulations. In this system, motion-onset visual evoked potential is used for characterinput, while motor imagery is used to switch the interface and allow to input the nextcharacter. Subjects can operate the whole system without the aid of any non-EEGsignals. Results of the experiments show that subjects can operate the system well afternecessary training, with a shortest average time of3.9s in single-step motor imageryoperation, and an accuracy up to93.3%in character input.
Keywords/Search Tags:Brain-computer interface, motor imagery, online, motion-onsetvisual evoked potential, hybrid BCI
PDF Full Text Request
Related items