Research On Recognition Of Depression Based On Facial Features

Depression,also known as depressive disorder,is characterized by significant and persistent depression.The prevalence and the disabled rate of depression is high,which has a serious negative impact o on the families and society of patients.The key to reduce the harm of depression is early diagnosis and treatment,so in recent years,researchers have focused on finding effective indicators which can objectively and quantitatively evaluate depression.The facial expressions of patients with depression are often characterized by sadness,depression,reduced smile and easy crying.Meanwhile,the collection of facial expression data has the advantages of easy access,non-contact and low cost.Therefore,the recognition of depression based on facial features has become a research hotspot in recent years.This paper focuses on the difference of facial expression between depressed patients and normal people under different stimulation conditions,and constructs four initial facial feature sets.Through feature selection and comparative analysis,we determine the effective features and build a classification model to improve the accuracy of depression recognition based on facial features.The main contributions and innovations of this paper are as follows:(1)It was found that the expression differences of depressed patients and normal people under different stimulation conditions were different,and the key location was determined.A special experiment was designed to identify depression.The facial data of the subjects were collected and the facial features were constructed.Through the comparative analysis of facial features,it was found that the main differences between depressed patients and normal people are the corners of mouth,cheeks and eyes,in which men are more likely to show differences in cheeks and eyes,while women are more likely to show differences in corners of mouth.(2)Based on the research results of(1),the facial feature set was simplified and the depression recognition model based on facial features was constructed.Cross validation was carried out on the face data sets of 160 subjects.The classification accuracy of male and female subjects can reach up to 81.4% and 80.0%,which verified the validity of depression recognition model.(3)The effect of experimental stimulus materials on the accuracy of depression recognition was studied.It was found that the most effective discrimination material was text reading.In addition,picture description and video watching can also stimulate facial expression changes.These conclusions provide valuable reference for the design of more targeted experiments.In this paper,the differences of facial expression between patients with depression and normal people are discussed in detail,various facial features are analyzed in depth,the effective features and key parts of depression recognition using facial features are determined,and a depression recognition model based on facial features is constructed.The validity of depression recognition model was proved by cross validation on 160 samples.The research results of this paper can promote the research of objective and quantitative evaluation of depression.