| As is well known,there are two important brain research fields,i.e.brain disease diagnosis and brain-inspired computing.The United States,European Union and China have already announced the ambitious brain projects independently.What we know about our brain actually is still very limited.Further researches are still required on attention,study,short-term memory,long-term memory,decision,cognitive and languages.We know the EEG has an excellent time resolution.Moreover,there are no risks associated with an EEG measurement The test is painless and safe.So it plays an important role on brain disease diagnosis and brain cognitive research.The EEG signal is generated by a great number of brain cells communicating with each other through electrical impulse.Usually there's limit number of participants in experiments to study cognitive,so it is always a challenge work to analyze EEG signals or extract the features.Recently,machine learning and deep learning show many excited results from different area.These algorithms can help us to study the brain functionalities.Certainly the progress of brain research often stimulates new algorithms.For example,the convolution neural network which successfully used in object recognition is loosely mimicking the way that a biological brain solves problems.Usually the cognitive experiments need to be done by a synchronized event or manually remove the artifacts,which impacts the algorithm be used in real-time applications,for example,virtual reality,brain-interfaces.The main work of this paper is as following.(1)In this paper,a new experiment was introduced to study the brain activities under attention task and visual stimulation.And we try to find an algorithm which need not manually remove artifacts.The experiment consists of three different conditions:idle with eyes open,idle with eyes closed and string recognition task.In nonlinear,we proposed symbolic sample entropy and construct variation topographies to analyze the EEG signals.Firstly we select four noise series to validate the new algorithm.The test series include brown noise,pink noise,blue noise and white noise.It shows that the symbolic sample entropy can perfectly detect the nonlinear correlation of the four noises.Then we apply the algorithm to the EEG signal recorded in the experiment.The variation topography highlights the differences at right hemisphere between string recognition and idle conditions,especially at leads P4,02,T6 and C4.SSE at right hemisphere,including the occipital,occipitotemporal,parietal,and central part significantly increases in string recognition tasks.In addition,we demonstrate that the results of the whole group are consistent by symbolic sample entropy,while the traditional sample entropy cannot get the consistent results across the group.Furthermore,we propose a new parameter of Normalized Rhythm Power(NRP)to analyze EEG signals.Compared with eyes open task,string recognition task has consistent elevation in gamma NRP at P4,02 and T6.The alpha NRP of eyes open is obviously lower than that in eyes closed state.As is known,the dominant visual pathway from retina to brain is the one which travels to primary visual cortex.When there are visual stimuli,the alpha desynchronization happens in visual cortex.It is shown in the variant topography based on symbolic sample entropy.At last,we have a discussion in both physiology and physics.Since symbolic sample entropy evaluates signals from the aspect of complexity and normalized rhythm power measurements evaluate signals from the aspect of frequency distributions,complementary information about the underlying dynamics can be provided through the two types of measurements.(2)In order to study the brain activities of work memory,we propose an algorithm to construct the brain functionality network with calibrated Pearson Correlation.The research is based on 16-lead synchronized scalp EEGs collected under three experimental conditions:quiet,control and memory task.We suggest calibrating the Pearson correlation in EEG analysis from two perspectives.On one hand,the Pearson correlation is only calculated for matched frequency bands.On the other hand,the correlation coefficient is compared with the one calculated from matched surrogates.Then two parameters are proposed in this paper.The first one is link strength,which is contributed only by those significant correlation coefficients compared with the value from surrogates.The second parameter is node connectivity,which is defined as the sum of all link strength linked with a node.With these two parameters,variation topographies are designed to analyze EEG signals.In beta and gamma rhythms,there are some profound results,(a)From the gamma topography,inter-hemisphere links are strengthening,while intra-hemisphere frontal-posterior links decrease.It implies parallel inter-hemisphere coordination as well as sequential intra-hemisphere frontal-posterior coordination in visual and motor processing;(b)Comparing the memory and control states with the quiet state,the beta topography indicates the links between T5/T6 and 01/02 are strengthen,which represents the visual ventral stream;(c)when comparing memory with control task,the gamma topography shows an increase in T6 node connectivity,which provides strong evidence for the information binding or relational processing function of the temporal lobe in memory forming from scalp EEG.To our knowledge,it is the first time to effectively capture brain connectivity variations associated with working memory from a relatively large scale both in time(from a second to a minute)and in space(from the scalp).(3)We proposed a symbolic joint entropy algorithm and rescale the value to a coupling coefficient with the range is[0,1].Instead of decomposing the EEG into different frequency bands(theta,alpha,beta,gamma etc),the novel algorithm introduced in this study is to compare the coupling strength from a whole frequency spectrum above 4Hz.We validate the new coupling coefficient on the well-known Henon map data firstly.Then the algorithm was applied to the EEG data from the attention cognitive experiment.The result shows that the coupling near occipital lobe in eyes close task is quite higher than the one in eyes open task.However,the result is not very consistent in all subjects.According to the neurophysiology,the alpha desynchronization usually happen at the back or posterior of the brain in eyes open task.But the symbolic joint entropy provided in the paper analyzes a wide frequency band above 4Hz.So it is not surprising that the result of symbolic joint entropy is not consistent.Furthermore,an important finding is that the coupling strength is obviously great in eyes closed state when there's some time delay applied,especially at P3,P4,T5 and T6,The analysis shows that the result is very consistent in the group.The algorithm we proposed not only can classify the two brain activities clearly,but also detects the physiology delay coupling characteristics. |
PDF Full Text Request