A Study On The Adaptive Method For Human-Machine Interaction Using Brain-Computer Interface

In recent years,although the Brain-Computer Interface(BCI)technique continues to develop rapidly,it is still limited by the information transfer rate and recognition accuracy.If the BCI technique can cooperate with other methods reasonably,some more complex control requirements could be achieved.The applicability of such systems could hence be further extended.On the other hand,the control methods based on eye tracking(ET)has been well developed.The main advantage of such methods is that the tracking of eye movement can be done quickly and accurately,and does not require any training of the users.Since the eye movement information itself contains only gaze positioning information,to determine whether the user is focusing on a certain position or a certain target,current ET based methods mostly compare the stagnation time of the user's gazing with a certain threshold.However,due to the complexity of various tasks,it is often difficult to select a suitable threshold for this.In this work,considering the advantages and disadvantages of both the BCI and ET technique,we try to use the eye movement information in positioning,and use the BCI technique to recognize the user's attention state directly.Hence,the two techniques can be integrated and cooperate in a newly proposed control system.This paper mainly demonstrates an EEG-based attention recognition method and a practicable scheme of the brain-eye-hybrid control.Our main works include:(1)Attention Recognition Based on EEG-To determine whether the user's concentration is focusing on something,a EEG-based method is proposed and implemented.Its operation process can be roughly divided into five stages:a.Signal Acquisition-collects EEG signals from multiple channels of the user using the SynAmps2 instrument;b.CSP Processing-the collected information is mapped via a(trained)CSP transformation matrix;c.AR processing-obtain the frequency spectrum of each channel from the CSP mapped signal;d.Feature Extraction-extracts spectral values in particular channels and at particular frequencies as recognition features,where the interested channels and frequencies are obtained by a regression analysis on a training dataset beforehand.e.SVM Classification-Using the(trained)SVM to classify the features,and the attention recognition is then achieved.This part of the work also includes training of the key models involved in the above process,as well as using an offline experiment to verify the validity of the method.(2)Practice of a brain-eye-hybrid control scheme-In this work,a system for controlling home appliances through a brain-eye-hybrid control technique was designed and implemented.The system provides a control interface for 3 kinds of home appliances,each of which contains 6 commonly used function buttons.Eye movement information is used as a cursor position in the system,and user's selection actions can be identified by online EEG-based attention recognition.At the end of this article,it can be seen from our experiments that the brain-eye-hybrid control can achieve good control effects.The control efficiency,the operation time cost,and the user experience are satisfactory.