Research On The Methods Of Brain-Computer Cooperative Control With Spontaneous Multi-class BCI

Brain-computer interface(BCI)provides a new communication and control channel between brain and computers,without the participation of peripheral nerves and muscles.The spontaneous BCI based on motor imagery(MI)has been widely applied on assistant devices control,due to its high response speed and autonomy.However,the number of control commands available in MI-BCI is still very limited,which restricts the usability of BCI systems in terms of multiple degrees of freedom(DOF)control applications.To address this problem,this dissertation presents a novel multiclass BCI design method based on sequential coding of MI tasks(sMI)to increase the number of BCI commands.The proposed sMI-BCI approaches were also used to control real assistant devices,such as humanoid robot,robot arm,and wheelchair,to demonstrate the practicability for multiDOF control.The contents of the dissertation are summarized as follows.Motor imagery sequential coding method for multi-class BCI design.To increase the number of BCI commands,this study proposed a novel motor imagery sequential coding(MISC)method for multi-class BCI design.Different brain activities were modulated by the sequences of MI tasks,which were constructed by left hand MI,right hand MI,and resting(no control motions).The sMI codes were detected from EEG and mapped onto special commands.According to permutation theory,an sMI task with N-length allows 2 ×(2N-1)selectable commands with the two MI tasks under self-paced conditions.To verify its feasibility,the MISC method is used to construct a six-class sMI-BCI paradigm.Four subjects participated in our experiments and the average accuracy of the six-class sMI tasks was 89.4%.The Coben’s kappa coefficient and the throughput of our BCI paradigm were 0.88±0.060 and 23.5 bits per minute(bpm),respectively.The above results demonstrated that the MISC method can effectively increase the number of BCI commands with reliable classification accuracy.Brain-actuated robot system using sMI-BCI paradigm.The sMI-BCI was applied to control a humanoid robot to test its usability for multi-DOF control.First,a two-class sMI-BCI was designed based on one class MI task to control the robot’s walk function.Four subjects participated in the experiments,and the average classification accuracy of the sMI tasks was 90.4%,with the average response time of 5.0 seconds.The average time of the four subjects completed one-walk task was 139.4 seconds by generating 11.9 commands.Then,a six-class sMI BCI was also constructed using left/right MI tasks,to control a multi-joints robot arm.The subjects completed one target grasp task in39.1 seconds on average,by generating 6.1 commands.The above results demonstrated the feasibility and practicality of the MISC method for multi-DOF control of real devices.Asynchronous sMI detection algorithm and its application in wheelchair control.To improve the autonomy of sMI-BCI system,an asynchronous detection algorithm for sMI tasks was presented in this study.The brain activities of starting,stopping,and switching over left and right MI tasks were detecting by a template matching method.The subjects execute the sMI task autonomously without interruption of the prompt messages of BCI system.A four-class asynchronous sMI-BCI paradigm was constructed using the above algorithm,achieving the average classification accuracy of 90.7%.The sMI-BCI paradigm was also used to control a wheelchair to evaluate its performance.All the subjects could perform the wheelchair control functions of left/right turning,starting/stopping,and ac/deceleration.The results demonstrated that the asynchronous detection algorithm of the sMI tasks can improve the asynchronous performance of sMI-BCI.sMI detection algorithm based on multi-scale window lengths.To speed up the detection of sMI tasks,a multi-scale window length algorithm was proposed in this study.Under this algorithm,eight different windows with the length of 600-2000 ms were selected to classify the MI EEG features.The confidence level of the classification results from the eight windows were calculated simultaneously,to determine the weights for the outputs of each pre-classifier.The final classification results of MI tasks were generated by the weighted results from the eight windows.Using the proposed algorithm,the response time of MI detection was decreased by 11.4%,comparing to the single window method.On the sample level of response time,the classification accuracy was increased by 4.2% by the multi-scale window algorithm.These results show that the proposed algorithm can speed up the detection of sMI tasks with reliable classification accuracy.The above studies demonstrated that the MISC method increased the number of MIBCI commands with reliable classification accuracy,which can be used to control the multi-DOF devices.This dissertation provided an effective method for multi-class MIBCI design,which is promising in multi-DOF control of assistant devices.