Attention State Detection Based On EEG

Attention refers to the ability of focusing people’s psychological activities such as perception and cognition on an object.Attention is critical to people’s daily life in multiple ways.Paying attention can significantly improve working and learning efficiency.On the other hand,distraction can lead to serious consequences under scenarios such as car driving.Attention can also be useful in medical treatment.Mental diseases,such as attention deficit hyperactivity disorder(ADHD),can be detected by analyzing the patient’s attention level.This thesis investigates an attention state detection algorithm based on EEG signal.The main work of this thesis includes:1.Using a portable dry electrode device to construct an Electroencephalogram(EEG)dataset under different attention states.Attention states are divided into three categories:focused,normal and unfocused,which correspond to three types of mental tasks:intensive mental arithmetic,daily work and rest.The dataset contains EEG signals collected from 40 people,and can be used for studies of attention state based on EEG.2.Proposing a feature extraction algorithm based on EEG attention state classification.This algorithm first uses wavelet de-noising algorithm based on self-adaptive threshold to preprocess EEG signals;the algorithm then extracts data features from four dimensions,namely time domain,frequency domain,timefrequency domain and entropy concept.The effectiveness of the proposed features is verified by experiments.3.Proposing a dual branch network structure to classify attention states based on EEG signals.This study constructs models with various functionalities by combining different modules.The first model is composed of logistic regression module and deep neural network module.This model is able to learn both the depth features of artificial features and deep network extraction,and obtains a higher classification accuracy than a single model.The convolution module based on time sequence can process mixed input signals composed of sequential signals and artificial features,and can accomplish the task of target classification better.The effectiveness of the proposed algorithm is verified on public data sets.4.Designing and implementing a demonstration system of attention detection algorithm based on EEG signal,which is capable of detecting attention state with low delay.