Research On Facial Emotion Recognition And Behavior Analysis Technology For Prisoners

In prison anti-investigation and non-cooperative confrontational environments,prisoners have the characteristics of strong concealment of behavior clues and serious psychological precautions.At present,most of the research on the reformation of prisoners is based on the normal methods of traditional questionnaires and interviews,as well as through physiological signals such as brain waves and heart rate.However,traditional questionnaire surveys and interviews have low accuracy and strong subjectivity,and it is difficult to mine the potential behavioral intentions and objective psychological activities of prisoners.Moreover,due to the presence of acquisition devices during the analysis and research based on physiological signals such as brain waves,it is easy for prisoners to have a high degree of vigilance and defenses.Therefore,the data obtained is not the most real and effective.However,facial expressions and behaviors as objective and non-contact information are of great use for the prisoners' emotion recognition and reformation assessment.In response to the problems above,this article will rely on the national key research and mainly relies on the national key R&D project of "Research on Reform and Evolution Correction Technology and Equipment for Prisoners".By analyzing the facial emotions and behaviors of prisoners,we then conclude whether there are negative emotions such as anxiety,depression,and anger and whether there are negative behaviors such as lying when talking to police officers and not cooperating with reformation.This can better help police officers analyze the reformation of prisoners.In terms of facial emotion recognition,an anxiety,depression,and anger emotion recognition techniques based on integral projection and DCP features are proposed.For the analysis of the facial behavior of prisoners,a micro-expression detection and lie detection algorithms based on combination of optical flow features and windmill mode features,as well as the prisoners' active reformation evaluation algorithm based on attention detection is proposed.Finally,a simple and easy-to-use facial emotion and behavior evaluation system for prisoners was developed.The main work and innovations of this article are as follows:? Negative emotions such as anxiety,depression,and anger have become important causes of psychological problems that endanger mental health.Due to the particularity of the prisoner,their mental health will directly affect their reformation status,and the three types of negative emotion facial expressions are very similar and difficult to distinguish.Therefore,an anxiety,depression,and angry expression recognition method based on integral projection and double-crossing mode is proposed,and a discrimination based on boundary Fisher analysis is introduced.By obtaining texture information of important facial expression areas,the overall projection that retains shape attributes of facial expressions are combined with cross-spatio-temporal texture attributes in a double cross pattern,to realizes anxiety,depression,and angry expression recognition.? The lie detection based on facial micro-expressions is to perform micro-expression detection first,that is,to detect micro-expressions in a video,and locate the positions of the start and end frames,and then determine whether there is appearance of lying after analyzing the key micro-expression.This paper proposes a micro-expression detection and lie detection algorithm based on the combination of optical flow features and windmill mode features.It divides the area of interest of the face according to different expression's FACS motion units;It proposes a micro-expression detection method combining optical flow features and windmill mode features,and visually detects micro-expression fragments by combining facial texture features and facial dynamic features,which greatly improved the performance of micro-expressions detection.Then the lie detection is realized based on the five key facial micro-expressions.? Research was done on the attention detection of prisoners when talking to police officers as an important basis for assessing whether prisoners are actively reforming.It integrates four types of facial behavior detection:face detection,head pose detection,eye closure detection and gaze estimation.We analyze the face images of the prisoners to determine whether they are focused or not,and use this as a basis to identify whether the prisoners are actively reforming.? The computer vision libraries OpenCV,OpenFace,and Python are used to construct the prisoner's emotion and behavior evaluation software system and a webcam is used for video capture.The three tasks are performed respectively:negative emotion recognition;micro-expression detection and lie detection analysis;negative emotion recognition;active reformation evaluation based on attention detection.We then display and analyze the results of each part.