Research On Active Brain-computer Interface Feature Recognition Based On Temporal Codin

Brain Computer Interface(BCI)is a new type of human-computer interaction system.It does not rely on the traditional way such as human muscle tissue,peripheral nerves to establish a new information exchange and control channel between human and the surrounding environment.With the BCI system,brain activity can be translated directly into the commands that drive the device.At present,there are two main types of BCI systems: One of them is evoked potentials based BCI systems,such as Steady-State Visual Evoked Potentials(SSVEP),P300 evoked potentials and emotion recognition BCI system,etc.Evoked potentials based BCI systems have the advantage of high classification precision.But the need for additional stimulation equipment,easy to cause fatigue;The other type is active BCI,such as motor imagery,speech imagery,etc.It does not need to impose additional stimulus,but the classification accuracy needs to be improved.Due to the limitation of the spatial resolution of Electroencephalogram(EEG).Motor imagery has relatively few operational dimensions,mainly including the left hand,right hand,feet,and tongue.In order to increase the number of active BCI instruction sets,this paper proposes an active BCI experimental paradigm based on the sequential coding of speech imagery and motor imagery.It mainly includes the binary experimental paradigm and the fourcategory experimental paradigm based on the sequential coding.On this basis,the feature extraction and classification of offline data are studied.The main contents of the study are as follows:For the binary experimental paradigm based on time sequence coding,its two kinds of imagery tasks are: 1)motor imagery;2)speech imagery first and then motor imagery.For the four-category experimental paradigm based on sequential coding,its four kinds of imagery tasks are: 1)motor imagery;2)speech imagery;3)motor imagery first and then speech imagery;4)speech imagery first and then motor imagery.For these two types of experimental paradigms,12 subjects were conducted offline experiments to collect EEG signals,and the feasibility of the proposed experimental paradigm based on timing coding was verified by ERD/ERS analysis of EEG signals.Aiming at the binary experimental paradigm based on sequential coding,a Sub-Time Window Filter Bank Common Spatial Pattern(STWFBCSP)algorithm is proposed.After dividing the imagery period into five time-windows,multi-band filtering is carried out for each time window,and then feature extraction is carried out by using Common Spatial Pattern(CSP),and the optimal feature is selected by using mutual information.The final decision output is determined by the voting results of five Support Vector Machines(SVMs).12 subjects participate in this experimental paradigm,and their average classification accuracy is 68.94%.Compared with CSP,FBCSP and CTFSP,the average classification accuracy of the proposed algorithm is improved by 14.38%,9.02% and 3.97%,respectively.The results show that STWFBCSP can better extract and classify the features of EEG signals based on the binary experimental paradigm based on sequential coding,so as to improve the practicability of BCI.Aiming at the four-category experimental paradigm based on sequential coding,after analyzing the temporal,frequency and spatial features of EEG signals,four types of signals are classified by a multi-classification model based on time sequence.According to the differences in time between the four types of imagery,feature extraction and classification are accomplished by CSP and SVM,respectively.12 subjects participate in this experimental paradigm,and their average classification accuracy is 68.94%.The classification results are much higher than random probability,so the proposed experimental paradigm is feasible and valuable.The experimental paradigm based on sequential coding can effectively increase the number of instruction sets of active BCIs,so the practicability of BCIs is also improved.In conclusion,this paper focuses on active BCI based on sequential coding of speech imagery and motor imagery,and feature extraction and classification algorithms of EEG signals are studied on the basis of off-line data analysis.It lays a foundation for the application of sequential coding based BCIs.