Research And Implementation Of Learning Acceleration System Based On Single Channel EEG

As time goes on,knowledge updating is more and more frequent.People must constantly absorb the nutrition of new knowledge to enrich themselves in order to adapt to the fierce competition of society.With the development of science and information technology,knowledge is already massive,but time is still limited.Therefore,more effective access to knowledge has become an urgent need for people.Based on this,this thesis studies the improvement of learning efficiency,builds a learning acceleration model,and correlates it with EEG signals.Finally,a learning acceleration prototype system based on single-channel EEG signals is designed and implemented.This thesis mainly completes the following work:(1)A learning acceleration model is established.We introduce the overall architecture of the learning acceleration model and describe its construction process.The relevant factors of learning state are screened,attention,fatigue,and mental workload are selected to represent learning state.Different EEG feature extraction schemes are proposed for the three representative factors of learning state,and the rationality is verified by experiments.A learning state assessment model is proposed and introduced in detail.(2)Intensive research on the processing methods of EEG signals.The related techniques in the EEG signal denoising method are introduced.The principles of the three denoising methods of stationary wavelet transform,EEMD-ICA and SCICA are described.According to the characteristics of EEG signals that are susceptible to eye movement and blinking during the learning process and the system's requirements for real-time computing,we compare and analyze the advantages and disadvantages of the above three denoising algorithms,and selecte stationary wavelet transform as the denoising method of EEG signals.The principles of variational modal decomposition and wavelet packet decomposition are introduced in detail,their decomposition efficiency and accuracy in the decomposition of real EEG signals are analyzed,and the more efficient wavelet packet decomposition is selected as the EEG signal decomposition method.The principles of KNearest Neighbor Classification,Support Vector Machine,and BP Neural Network are described in detail.Their classification effect of the representative factors of learning states is analyzed.Finally,the support vector machine is selected as the classification method for learning state representative factors.(3)A learning acceleration prototype system based on single-channel EEG signals is implemented.Combining learning state screening,EEG signal processing algorithms and learning state assessment model,a learning acceleration prototype system based on single-channel EEG signals is realized.The prototype system can evaluate learners' learning state and provide corresponding suggestions.At the end of the thesis,we test the learning acceleration prototype system,and design different tests for three learning state representative factors of attention,fatigue,and mental workload,and analyze their classification accuracy.In addition,the thesis tests the validity of the learning state assessment of the learning acceleration prototype system.The test results show that the learning acceleration prototype system realized in this thesis can effectively identify learners' learning state and provide appropriate suggestions.