Electroencephalogram Data Analysis And Brain-Inspired Intelligence Obstacle Avoidance Strategy About BCI System

With the rapid development of computer science and the continuous extension of brain neuroscience,brain-computer interface system(BCI),as an effective brain-information exchange channel which provided a proper communication environment between human thinking and external devices,has attracted more and more researchers.Through the BCI system,people have created an innovative human-computer interaction method based on human thinking.BCI technology has been applied to auxiliary rehabilitation,traffic control,military strategy,entertainment and some other aspect,which shows a great research value.In this paper,we studied the related technology about BCI system which including the de-noising of EEG,the recognition of EEG and the obstacle avoidance strategy of brain control robot.The details are as follows:It is the first key point to cancel the noises of EEG for the BCI applications.The traditional de-noising method mainly involves the EEMD algorithm.As the noises are always found in high frequency domain,the EEMD algorithm often remove the high frequency IMF components directly to realize the de-noising.Although the EEMD can realize the de-noising,it will always remove some effective information at the same time,which will have a serious impact on the recognition of EEG.This study then proposed an innovative DTCWT-EEMD de-noising algorithm to make a better performance.For the DTCWT-EEMD algorithm,the first step is decomposing the EEG signals with EEMD algorithm,the second step is transferring the high frequency IMF components which contain the noises with DTCWT algorithm,in the next moment,we will make a soft-threshold de-noising for the wavelet coefficient,and in the end,we will reconstruct all components and get the clean EEG signals.So,the DTCWT-EEMD de-noising algorithm can realize the better de-noising,and meanwhile,it can retain the useful information as far as possible to ensure the completeness and validity of EEG signals.The second key point is the recognition of EEG.At present,for the process of recognition,the traditional method often split the feature extraction and the pattern recognition.Although some traditional methods can solve the recognition of EEG,for some complicated EEG,it still has some problems such as the intricate steps and the classification accuracy.In order to solve the above problem about motor imagery EEG,this paper proposed an improved stacked de-noising auto-encoders network combined with the immune optimization algorithm and decision making mechanism(ISDAE).The ISDAE network extracted the most effective feature vectors by multi-layer DAE,and then identified these feature vectors by the neural network.And meanwhile,the decision-making mechanism was obtained and the immune optimization algorithm was used to optimize parameters of ISDAE,which results in the improved ISDAE model with higher classification accuracy.The experimental results show that the ISDAE has a strong ability for feature learning from raw EEG data and a higher recognition rate,which provides an efficient technique for the recognition of motor imagery EEG.The third key point is about the strategy of brain control.In order to solve the problem about the lower security and stability for the brain-controlled robot movement control,this paper proposed an intelligent collision detection model based on visual brain mechanism.By simulating how the brain-visual system calculates the time to collision(TTC),the proposed intelligent collision detection model adopts the spiking neural network to extract the moving target and acquire some information about the angle of view,and used the back-propagation neural network to predict and calculate the TTC,and meanwhile,the proposed model imitate the brain memory control with the MAP structure.The experimental results show that the intelligent collision detection model based on visual brain mechanism can achieve the accurate prediction about the approaching obstacle's TTC,and we also can get a better performance about dynamic obstacle avoidance by using the proposed model.So,the intelligent collision detection model provides a crucial foundation for the obstacle avoidance movement and enhances the safety and stability for the brain-controlled robot.In the end,a summary analysis about related research contents was carried out in this paper.And meanwhile,the future research direction is given out.