Research On Key Technologies Of MI-BCI Based On Visual-auditory-somatic Assisted Stimulus Feature Modulation

Brain-computer interface(BCI)interacts information through cerebral activities and external environment.Its composition usually includes a series of hardware devices and software technologies such as signal acquisition,processing and output.Motor imagery(MI)is one of the common paradigms of BCI,which means that the brain has the intention of action but without performing the actual limb movements.MI-BCI can be used to repair brain plasticity in patients with motor function injury.However,there are some problems to be solved in conventional MI-BCI,such as low decoding accuracy of motion intention,poor spatial resolution,large individual difference of imagination ability,“BCI blindness” and less imagery limb tasks.Studies have shown that the usage of assisted stimulus in MI-BCI can effectively improve the BCI performance,but it is necessary to design an efficient assisted stimulus system.In order to solve the above problems,this thesis proposes a new visual-auditory-somatic assisted stimulus method based on feature modulation.A series of work around the key technologies in BCI are carried out,such as design of new system paradigms,the feature modulation patterns of the assisted stimulus,the cerebral neural response process as well as the decoding accuracy of motion intention.The main research work includes:1)A feature modulation method of motor imagery based on dynamic visual assisted stimulus is introduced in this thesis.The brain oscillatory patterns,feature modulation effect and BCI motion intention decoding performance under different visual assisted stimulus are studied.It is confirmed that the recognizable features of complex MI tasks combined with dynamic visual stimulus are significantly enhanced.The average classification accuracy of the system is 14% higher than that of the static paradigm,and the individual optimal accuracies are all more than 87.5%.This study could provide a new way for the application of MI-BCI in motor function rehabilitation.2)The information processing and transmission rules between sensorimotor cortex and other cerebral cortex under visual assisted stimulus are discussed.The cortical current source densities under dynamic stimulus combined with complex MI task are significantly stronger in motor related areas and parietal frontal mirror areas.The causal connections in brain network are more complex under dynamic stimulus with complex MI task,especially in supplementary motor cortex and parietal mirror cortex.This study provides an important basis for revealing the modulation patterns of assisted stimulus and the way of brain neural response.3)A new method based on dynamic auditory stimulus is put forward to meet the needs of people with visual impairment.The modulation effect of dynamic auditory stimulus on brain nerve activity during MI task is verified through the analysis of timespace-frequency features.The decoding performance under dynamic auditory stimulus can meet the requirements of BCI application.4)A new hybrid feature enhancement method combining steady state somatosensory assistant stimulus and lower limb MI is studied.It is confirmed that MI task has modulation effect on somatosensory features.After using the hybrid feature decoding strategy,the recognition accuracy of left or right foot motion intention is improved by 16%.This work can solve the problems of low spatial resolution of lateral recognition of foot MI task and less single limb action command,which could expand the application of MI-BCI in the field of lower limb motor function rehabilitation.