Open AccessDynamics of Key Facial Points as an Indicator of the Credibility of Reported Information
Author(s) -
V А Barabanschikov,
А В Жегалло
Publication year - 2021
Publication title -
èksperimentalʹnaâ psihologiâ
Language(s) - English
Resource type - Journals
eISSN - 2311-7036
pISSN - 2072-7593
DOI - 10.17759/exppsy.2021140207
Subject(s) - computer science , principal component analysis , credibility , set (abstract data type) , face (sociological concept) , key (lock) , artificial intelligence , dynamics (music) , interval (graph theory) , component (thermodynamics) , data mining , pattern recognition (psychology) , mathematics , psychology , social science , computer security , combinatorics , sociology , political science , law , programming language , pedagogy , physics , thermodynamics
This research describes a method for studying the authenticity/unauthenticity of the information re- ported by people in video images. It is based on automatic tracking of coordinates of key points of a speaker’s face using OpenFace software. When processing the data, the multiple linear regression procedure is used. It was found that the dynamics of neighboring key points in the obtained models has a multidirectional char- acter, indicating the presence of a superposition of several dynamic structures, corresponding to the characteristic complex changes in the face position and facial expressions of the sitter. Their isolation is realized by means of the principal component analysis. It is shown, that the first 11 principal components describe 99.7% of the variability of the initial data. The correlation analysis between the number of credibility/confidence statements on the set of time intervals and the principal component loadings, allows to differentiate the dynamic structures of the face, connected with the assessments of credibility of the reported information. Automated analysis of face dynamics optimizes the process of collecting empirical data on the sitter’s appearance and their semantic structuring, as well as expands the range of predictors of the assessments of the truthfulness of the messages received.