Research And Application Of Mindfulness Awareness Attention Model Based On EEG

COVID-19 has had a profound impact on people’s lifestyles,social relationships and work status.Many people have been forced to stop working,lose their jobs or require home quarantine because of the COVID-19,and these changes can lead to negative emotions such as extreme depression,anxiety and frustration.Numerous studies have shown that mindfulness meditation is a skill that develops self-awareness and attention to the present through concentration and self-observation.It has been shown to be effective in reducing depressive symptoms,improving emotional stability,reducing anxiety and enhancing cognitive abilities.However,most of the current research has focused on the use of mindfulness meditation in people with depression or recurrent depression,and the analysis of EEG signals in these people,or the use of neurofeedback in combination with mindfulness meditation,and few studies have actually tested and assessed level of mindfulness.The level of mindfulness indicates a psychological state of acceptance of mindfulness.By testing the level of mindfulness,on the one hand,individual’s ability to mindfulness and cope with depression and emotions can be understood.On the other hand,the effectiveness of interventions can be assessed.This thesis can understand changes in the individual’s mindfulness and adjust intervention programmes and strategies in time.People with low level of mindfulness usually have poorer emotional regulation and higher level of depression and anxiety,and are therefore more likely to develop depression and anxiety disorders.This is why it is essential to test one’s own level of mindfulness.Therefore,this thesis conduct an experiment to detect the level of mindfulness in people with subthreshold depression,and construct an EEG-based Mindfulness Attention Awareness Model to detect the level of mindfulness,and developed a visualisation system that can implement data reading,feature extraction,feature selection and identification and evaluation of level of mindfulness.The main contributions and findings of this thesis are shown below:(1)A Feature Weighting Based on F1＿weighted Score(FWBFWS)method is proposed.In order to detect and evaluate the level of mindfulness,this thesis conduct experiments to detect the level of mindfulness,collects EEG signals from two channels,Fp1 and Fp2,and conducts multiple sets of experiments on four classifiers by four feature selection algorithms,and finally selects 36 EEG features extracted by the mutual information-based feature selection algorithm as the initial feature matrix of this model.Then,in order to assess that each feature has its own importance,A Feature Weighting Based on F1＿weighted Score is proposed,and the experimental results of different feature weighting methods on the four classifiers are compared on the public datasets and the dataset of this thesis,and the 36 feature matrix weighted by the FWBFWS method are finally selected as the input to final feature matrix.(2)The Mindfulness Awareness Attention Model based on EEG(EEG-MAAM)is proposed,which combines the Electroencephalogram(EEG)and the Mindfulness Awareness Attention Scale(MAAS).Mindfulness Awareness Attention Scale(MAAS)is the current mainstream instrument for assessing levels of mindfulness.The model consists of 5 steps,namely data collection,data pre-processing,feature selection and feature weighting,model training and model evaluation.The model is also constructed using the 36 EEG features calculated in the previous section as input,which is then trained using Deep Belief Networks(DBN),replacing the top-level Softmax classifier with Random Forest(RF)classifier to construct a DBN-RF classification model,while comparing the other five classification models to detect the level of mindfulness in the constructed dataset of this article.The results using several evaluation metrics show that the DBN-RF classification model can achieve a maximum classification accuracy of 91.23%,Recall of 92.51%,92.50% and 89.64% for low,medium and strong levels of mindfulness,and F1 Score of 94.51% and 90.68% for medium and strong levels of mindfulness.Compared with the other five models,it proves the effectiveness of this thesis’ s classification model and further proves that this thesis’ s EEG-MAAM model has usability.(3)A system for detecting and evaluating the level of mindfulness based on the EEG-MAAM model is developed.The system is mainly implemented in Java,with some functions based on Python and Matlab.The main purpose of the system is the detection and identification of the level of mindfulness,and the main modules include data reading module,feature extraction module,feature selection module and identification and evaluation of level of mindfulness module.A practical application of the model proposed in this thesis is carried out,as well as a preliminary system visualisation of the work in this thesis.