Research On Emotional Modeling Of Hand-drawn Lines

In Chinese and Western painting and calligraphy,artists can convey a wealth of emotion to the viewer through lines.In response to these phenomena,this study chooses to take a computational perspective as a starting point,going beyond the usual approach of art theory,with the aim of exploring the computability of the phenomenon of hand-drawn lines conveying emotion and revealing its inherent laws.The key works of this paper are as follows:(1)This study first established the experimental framework of emotional transmission phenomenon of hand-drawn lines and the quantitative method of each link.The emotions were quantified as 28-dimensional discrete emotional words and the hand-drawn lines were quantified as a composite data form consisting of static image data and motion time-series data.The aim of this operation is to ensure that the experimental framework and the quantification method are reasonable,taking into account the influence of the motion factors of the hand-drawn lines on the expression of emotion.On this basis,a classroom pre-experiment was conducted to verify the feasibility of the framework and quantification method.(2)Next,the study measured the emotional transmission phenomenon of hand-drawn lines and created a dataset of hand-drawn lines with emotional labels.Specifically,the study collected motion time-series data of the test subjects when they drew lines expressing specific emotion words based on a digital screen and a self-developed system.Furthermore,this study measured observers’ perception behavior and augmented each sample with Russell’s two-dimensional emotional space coordinates and emotional transmission ability score labels within the emotional words,based on the existing discrete emotional word labels.This study resulted in a dataset of 1106 original samples and 4424 sample data,each with three emotional labels,which can provide strong support for subsequent analysis.3)Based on this,this study developed an interpretable model between lines and emotions.We first extracted the widest range of line features to date,including not only static features but also a large number of motion time-domain and frequency-domain features.Up to 3695 features were extracted based on a single sample,resulting in a feature set containing 4086670 items of data.Next,a Gaussian mixture model was used to cluster the Russell’s two-dimensional emotional space coordinates labels of the samples into seven ordered classes in the pleasure and arousal dimensions,respectively.The original feature set was then reduced in dimensionality using distance correlation coefficients,random forests and mutual information feature filtering,and a multinomial ordered logistic regression was used to establish interpretable models between the reduced feature set and the ordered ranks of the pleasure and arousal dimensions.The models pass the test well,verifying the computability of hand-drawn lines expressing emotional behaviour in the emotional transmission phenomenon of hand-drawn lines,and the strong interpretability of the models reveals many potential patterns behind this phenomenon.(4)Finally,this study further explored the phenomenon of perceptual differences in different forms of samples.By calculating emotion transfer indicators and transforming PCA into Russell’s two-dimensional emotional space,the differences and patterns of emotion transfer of different forms of samples were explored.The above study found that the emotional transmission phenomenon of hand-drawn lines is computable.The interpretable models developed in this study reveal several potential laws that are complementary to the field of line emotion metaphors.At the same time,the dataset developed and the important features identified in this study are potentially valuable in developing computational applications of line-related emotions.The experimental framework and quantification approach of this study takes full account of the motion factors of lines,providing a more precise and detailed means of quantifying their emotional expression.The exploration of different forms of samples contributes to an in-depth understanding of the phenomenon of perceptual differences in the emotional transmission of lines.These findings and results help to develop the application of line-related affective computing and contribute to the development of the field.