Research On Eeg Intention Recognition Of Dynamic Complex Object Control Based On Discriminant Analysis

Brain Computer Interface(BCI)is a new interactive combination of engineering technology and neuroscience.It can be used as a neurological rehabilitation tool to improve the motor or cognitive abilities of patients with neurological diseases such as stroke and quadriplegia.BCI based on motor imagery(MI)can reflect the subject's motor consciousness without external stimulation,which has played an important role in the induction of brain plasticity.It has theoretical research value to improve the accuracy of Electroencephalogram(EEG)recognition in the study of motor intention recognition of EEG.However,the current research based on MI-BCI still has some shortcomings,such as the low complexity of action paradigm restricts the quality of evoked EEG and the performance of motor intention decoding needs to be improved.In view of the above problems,the paper has carried out the following work:(1)We propose a novel task of visual guidance for controlling complex constrained objects.The common real-life scene of moving a cup of water to the table no liquid being spilled is simulated.And a conceptual model of the dynamic complex system is abstracted.Theoretical analysis is carried out from the perspective of dynamic critical energy.The mathematical model of the controlled object is established according to the Euler Lagrange equation and balance equation,and the virtual task scene is programmed.(2)Linear discriminant and quadratic discriminant analysis(QDA)are used to decode the EEG intention of constrained objects.Based on QDA,a nonparametric generalized quadratic discriminant analysis algorithm is proposed.By introducing the threshold c,the parameters can be adaptively adjusted to adapt to the data to be classified.The improved algorithm is applied to EEG signals in complex coupled object manipulation tasks to identify the subject's movement intention.Furthermore,regularized discriminant is used to analyze EEG intention,which improves the influence of multicollinearity by modifying the singular covariance difference.Compared with other classification methods,it is found that generalized quadratic discriminant analysis can significantly enhance the performance of motion intention decoding.(3)In decoding research,we discuss the effects of concentration and emotion on EEG.The decoding of motor intention in two different situations of desired and undesired feedback under complex coupled object operation tasks is studied.In the case of undesired feedback,the area under the curve increased significantly,which was statistically tested by page trend test,p < 0.05.The differences of brain activity under desired and undesired feedback tasks are analyzed using brain-network-based topographical scalp maps.Furthermore,it reveals the effect of task concentration on the performance of EEG.The results show that the curve accords with the Yerkes-Dodson law in psychology,subject's performance increases as levels of arousal become high.This paper makes an in-depth analysis and rese arch on the novel EEG evoked task from multiple perspectives,which provides theoretical support for high-precision analysis of motor intention and deep exploration of brain cognitive function.