Evaluation Of Personality Traits Based On The Fusion Of Temporal And Spatial Features Of Handwriting

Handwriting refers to the traces left by human beings in the process of writing,which is the product of the interaction and mutual influence between the physiological and psychological mechanisms of the writers.The brain is the basis for the formation of handwriting,and writing is a reflex activity of the brain.The individual writing process needs the coordination of the brain,cerebellum and body trunk,and will be affected by psychological factors such as emotion and personality,among which personality is the deep factor affecting handwriting.The relationship between handwriting and personality is so strong that a writer’s handwriting can be used to analyze his or her personality traits.The research of handwriting psychology covers different disciplines such as psychology,cognition,computer science,etc.The analysis of handwriting psychology is also widely used in criminal investigation,management.psychology and education,etc.At present,the traditional machine learning method and factor analysis method in statistics are mostly used in the study,of handwriting psychology,while there are few studies using deep learning to analyze handwriting psychology.The traditional machine learning method and factor analysis method both require manual selection of handwriting features,but there is no unified standard for the selected features,and the analysis of handwriting psychology is not universal.In other studies,although deep learning methods are used to study handwriting psychology,they only stay at the stage of handwriting recognition and fail to study the internal relationship between handwriting and psychological characteristics in depth.To solve the above problems,this paper proposed a handwriting and the space-time characteristics based fusion method of personality assessment.The proposed method is intended to study handwriting characteristics and psychological characteristics of writing the mapping relationship between personality traits,as well as construct a computerized quantitative model between handwriting characteristics and 16PF personality factors,make the psychological analysis of handwriting more universality,promote the handwriting psychological practice application in various fields.The main work of this paper is summarized as follows:(1)This article first uses two methods to expand the data set.one is the generation of adversarial network variant WGAN-GP,and the other is data expansion based on stroke segmentation.Then these two data sets are used to study handwritten Chinese characters and personality traits.And construct a computerized quantitative model between handwriting characteristics and 16PF personality factors,and propose a personality trait evaluation method based on handwritten picture characteristics.In this paper,a regression model based on ResNet is constructed to extract handwriting-related features in pictures of handwritten Chinese characters.The output of the model is 16 predicted scores corresponding to the 16PF personality factor.The MSE on the data set after the two data expansions reached 2.5604 and 2.4123,and the MAE reached 1.1787 and 1.1493,respectively.It shows that the proposed method can evaluate the personality traits of the writer through the handwriting to a certain extent.(2)Improve the proposed personality feature evaluation method based on handwritten picture features,and propose a spatiotemporal feature fusion model based on ResNet and BiLSTM.Among them,the ResNet network is used to extract the spatial features of handwritten Chinese character images,and the BiLSTM is used to extract the time series features of handwriting,and the Attention mechanism is used to assign different probability weights to the stroke time series features at different moments in the handwriting.Then we build a spatiotemporal feature fusion module,and used regression methods to predict the scores of 16 personality factors for handwriting features.Experiments on two handwritten Chinese character datasets prove that the performance of this method is better than that of the personality assessment method based on the characteristics of handwritten pictures.