| EEG has become a hot spot in the research field of brain-computer interface because it can truly reflect human brain activity.In various research based on EEG,attention detection has been widely concerned because it is closely related to people’s life.In this paper,an algorithm based on Synchronous Brain Network(SBN)is proposed,which uses EEG signal to calculate the Phase Locking Value between two electrodes.Convolutional Neural Network(CNN)is selected as a classifier because it can better extract the structure information of the Network and find the important features related to attention through self-learning.In order to further improve the classification accuracy of the algorithm and make up for the defect that the SBN algorithm fails to extract signal time domain information well,an attention recognition algorithm based on Synchronous Brain Network and Long Short-term Memory(SBN-LSTM)is proposed.Based on the SBN algorithm,the convolutional neural network structure of the SBN algorithm is used to extract the sample structure information,and the time domain information extracted from multiple time series samples is then extracted and integrated through Long Short-Term Memory(LSTM).And finally output the classification results.Compared with SBN algorithm,this algorithm is more suitable for current application scenarios because it can simultaneously extract time domain and structure information from samples.Experimental results also show that the classification accuracy of this algorithm is slightly higher than that of SBN algorithm.Aiming at the problem of poor practicability caused by long calculation time of SBN algorithm,an attention recognition algorithm based on Synchronous Brain Network and Information Gain(SBNIG)is proposed.Based on SBN algorithm,Added a feature sparse process based on information gain,the input of the characteristic matrix classifier needs an sparse before operation,so as to eliminate redundant features for classification provide less contribution,keep high contribution characteristics,under the premise that possible to ensure accuracy of classification,so both reduces the computing PLV value and the time needed for classification results,The efficiency of the algorithm is greatly improved,and the number of electrodes needed for classification is reduced,which is beneficial to the application of the algorithm. |
