Enhancement And Recognition Of Electroencephalography Features Induced By Movement Intention For Brain-Computer Interfaces

Movement is one of the essential functions of human daily life.The immense repertoire of motions from the activity of muscles are under the control of the nervous system.In recent years,the brain-computer interfaces(BCIs)driven by neural activities give their users communication and control channels that do not depend on the brain’s normal output channels of peripheral nerves and muscles.Notably,movement intention based BCIs have important scientific significance and applied value in motor rehabilitation,replacement and enhancement,which have attracted the attentions of scientists.The movement intention includes movement preparation before the movement execution and the motor imagery without actual body movements,which is confronted with several challenges in feature recognition.These challenges due to the short neural response time,significant individual differences and weak motor evoked patterns.This paper conducted deep researches on the enhancement of lateralization electroencephalography(EEG)features in time and frequency domains,indicidual resting state EEG features and the force load evoked EEG features,aiming to solve the problems in this fieldFirst,to solve the recognition problem of movement preparation,this paper extracted the pre-movement EEG signals induced by voluntary finger movement.By analyzing the time and frequency features of EEG,we could find that the movementrelated cortical potential(MRCP)and event-related desynchronization(ERD)features were complementary to each other.We designed a novel compound spatial filter algorithm which was based on discriminative spatial pattern and task-related component analysis for feature extraction of the low frequency MRCPs.Furthermore,based on the complementarity of the MRCP and ERD features,we proposed a feature extraction and fusion method for the multi-dimensional time-frequency-space features evoked by motor preparation,which utilized fisher discriminant analysis(FDA)classifiers.The average accuracy was 83.04%.An open dataset was used to verify the validity of the algorithm.The results showed the feature extraction and recognition method proposed in this paper was superior to the existing literature reports.Second,to evaluate the motor imagery evoked neural responses with individual differences,a new nonlinear dynamic neurophysiological index was proposed.Then,this study analyzed the correlation between the resting state EEG features and the motor imagery performance of three different paradigms across 105 subjects.The resting state EEG features included power characteristics and nonlinear dynamics characteristics in different frequency bands.The results showed that the alpha band in the open eye resting state was the key spontaneous neural oscillation rhythm which affected the motor imagery evoked neural responses.Furthermore,screening models were built.The research obove results provide a theoretical basis for further developing the new neurophysiological indexes related to motor imagery performance and establishing a robust screening model.Last,in terms of extending the limited command set of motor imagery,this paper proposed a novel BCI paradigm based on multi-force motor imagery.The subjects could drive an online 3 force motor imagery related commands BCI system using the same limb for the first time.The results showed an average accuracy of 70.9% for three classes,with the highest accuracy of 83.3%,which demonstrated the feasibility of the paradigm.This paper also analyzed the of neural oscillation patterns induced by the different force motor imagery tasks.A stronger activation on the sensorimotor area could be observed under the high force motor imagery task.At the same time,it showed the complementarity of the time and frequency features.This proposed paradigm could not only be used to expand the command set and application,but also be helpful for stroke rehabilitation.Given the above,this paper focused on the enhancement and recognition for the key features of EEG response evoked by movement intention in the BCI.This paper researched neural oscillation patterns induced by motor intention,proposed the compound spatial filtering and multi-dimensional feature extraction methods,analyzed the correlation between the resting state EEG and the subjects’ neural responses of motor imagery,and developed a novel multi-force motor imagery based BCI system,which might provide the theoretical basis and technical support for efficient braincomputer interaction based on movement intention.