Research On Influence Of Indoor Thermal Environment Transients Onlearning Attention

The recent development of IOT witnessed that the smart campus and classroom,applied in the new educational modes and teaching platforms,have been of increasing significance in the fields of education section.From the perspective of intelligence of smart classroom,the indoor thermal environment draws much attention in the construction of smart classroom,in that the not only does the indoor thermal environment directly affect thermal comfort of learners,but significantly relates to learning efficiency,psychological and physical health.Simultaneously,the evaluation of the indoor thermal environment's influence on learners' concentration is also considered as the basis for establishing the quantitative relationship between indoor thermal environment and learners' heed.Therefore,the thesis aims to explore the effect of the transient thermal environment on the learners' heed to quantify the learners' attention span.The participants in the current experiments were assigned to take part in the selected MOOC courses in two different thermal comfort levels,from which the indoor environment parameters were collected.Three physiological signals including electroencephalogram,electrocardiogram and skin electricity were collected by BP electroencephalogram equipment and multi-channel physiology apparatus.Moreover,the Profile of Mood Stares-Short Form questionnaires were collected to investigate their perceptions.The reliability and validity of the Profile of Mood Stares-Short Form questionnaire suggest that the data in the experiment is credible and applicable.The collected EEG,ECG,and GSR signals were preprocessed and feature-extracted respectively,and the linear and nonlinear characteristics of each physiological parameter were analyzed.Decomposing and denoising EEG were based on the db4 wavelet in 4 layers,which extracted approximate entropy and energy of EEG as nonlinear features.The db5 wavelet was used to decompose and denoise ECG in 6 layers,and four features such as the average heart rate and R_R interval of the ECG were extracted.The db5 wavelet was adopted to decompose and denoise GSR in 2 layers,contributing to the extraction of four linear features of GSR,for instance,standard deviation.In this research,we present the first evidence that approximate entropy and energy features of participants' EEG signals changed over time.K-means clustering algorithm was used to cluster the approximate entropy and energy features of EEG extracted.According to the clustering results,participants' concentration levels began to improve with the increasing thermal comfort after experiencing transient indoor thermal environment.Besides,it appeared that attention levels began to leap and centralize at the criteria of around 200 s.Moreover,in order to explore the relationship between ECG,GSR and learning concentration,the BP neural network was employed as a classifier,together with the result of the clustering of EEG used as label to classify the features extracted from ECG and GSR.The results demonstrated that the average correct rate of feature classification of ECG and GSR after classification by BP neural network was 89.47%,and the area under curve of the classifier's ROC is 0.89.The results suggested that ECG and GSR are able to contribute to the classification and evaluation of learning attention.