Classification Of Aesthetic Evaluation Of Chinese Typeface Based On Multi-domain Features Of EEG

Chinese typeface,one of the most important attributes of Chinese characters,widely exists in people's daily life.The aesthetic evaluation model of Chinese typeface can not only assist the typeface designer in the aesthetic design process of Chinese typeface,but also intuitively give the learners aesthetic feedback in the process of learning Chinese calligraphy.It has important value of both commercial application and art.Most of the existing computer aesthetic evaluation focuses on images such as photos,paintings,web pages and logos.However,there are few studies on the aesthetic evaluation of Chinese typeface.The research of computer image aesthetic evaluation generally extracts visual aesthetic features from the images,and then trains classifiers or regressors to fit the results of human aesthetic evaluation of the images.It is an objective aesthetic evaluation method driven by image data,ignoring the subjectivity and difference of human aesthetic cognition.Therefore,based on the difference of human's aesthetic preference for Chinese typefaces in brain neural activities,this paper uses the Electroencephalography(EEG)data of 17 subjects'aesthetic preference for Chinese typefaces and deep learning to explore the classification method of Chinese typefaces' aesthetic evaluation.Overall,This paper proposes a classification model of aesthetic evaluation of Chinese typefaces based-on multi-domain features of EEG.The work of this paper is mainly divided into two parts.The first part is EEG feature extraction.In view of the non-stationary characteristics of EEG,this paper uses continuous wavelet transform(CWT)to extract the time-frequency domain features of specific frequencies in EEG.In order to solve the problem that the spatial information of the electrodes can not be fully utilized by simply connecting the features of multiple domains of EEG to form the feature vector,this paper maps the three-dimensional spatial coordinates of 62 electrodes and the corresponding time-frequency domain features to generate the EEG feature map.It preserves both the space and frequency domain features of EEG.The second part is the construction of EEG classification model.First of all,according to the characteristics of EEG feature map,this paper designs convolutional neural network(CNN)classification model,which analyzes the performance based on different features of frequency combination and different time windows.Secondly,in order to solve the problem that CNN can not make full use of the temporal information in EEG,this paper proposes a EEG classification model,which integrates the time,space and frequency domain features of EEG.The model consists of two sub-networks:multiple time window CNNs and long short term memory(LSTM).By using feature splicing and fusion strategy,CNNs network which focuses on spatial and spectral information and LSTM network which focuses on temporal information are fused to learn the temporal,spatial and spectral fusion features of EEG and realize high-precision EEG signal classification.The effectiveness of this method is verified by comparing the effects of different time window combinations on model classification performance.