Research On Construction Of Depression Recognition Model Based On Multi-source EEG Data Fusion

In recent years,with the increasing incidence of depression worldwide,the diagnosis of depression in various fields has become a hot topic.At present,the diagnosis method of depression still adopts the subjective identification method,which will cause the missed judgment and misjudgment of depression.In order to find an objective diagnostic method,a large number of researchers are trying to find a breakthrough in the physiological signals or behavioral signals of depression patients,and relevant research is also gradually carried out.EEG,as a typical physiological indicator,has been used in the diagnosis of depression by some researchers and has achieved good results.However,as the underlying physiological mechanism of depression is still unclear,the model of depression recognition based on single-source EEG data has some limitations and is difficult to be applied universally.Therefore,in order to explore the pervasive EEG recognition model of depression,this paper fuses the EEG data of depression under different types of emotional stimuli at the feature layer by combining the multi-source data fusion technology,so as to build a relatively effective and well-generalized depression recognition model.The main contributions and innovations of this paper are as follows:(1)Using of depression in patients with the characteristics of the different reactions to the emotional stimuli,starting from the three kinds of emotional stimuli,including positive,negative and neutral,extracting the EEG feature associated with depression under Fp1,Fp2,Fpz three electrodes and under theta band,alpha band,beta band,gamma band and all band wave five bands.Based on the extracted features,45 singlesource feature subsets are constructed.(2)Aiming at the problem of uneven distribution of individual classifier weights in multi-classifier combination learning and the blindness of initializing population in genetic algorithm,a multi-classifier combination strategy based on improved genetic algorithm is proposed.The combination strategy was verified in 18 public data sets.The results show that the combination strategy is practical and effective.In addition,this paper uses this combination strategy to construct a multi-classifier combined learning model,and uses this model to select 5 relatively good data subsets from 45 single-source depression EEG data subsets as the basis of multi-source EEG data fusion.(3)Based on the analysis of the traditional multi-source data feature layer fusion strategy,a feature layer fusion strategy based on weight optimization is proposed,and this strategy is compared with the traditional fusion strategy.The results show that the fusion strategy is effective in depression recognition.In addition,in order to test the effectiveness of the depression recognition model based on multi-source EEG data,this paper compares it with a model based on single-source EEG data.The experimental results show that depression recognition model constructed based on multi-source EEG data can achieve better recognition accuracy,and to a certain extent can provide an auxiliary tool for the diagnosis of depression.