Research Of Depression Recognition Based On Expression Behavior Pattern

Depression is a common mental illness.According to data released by the World Health Organization in 2017,there are more than 300 million people with depression in the world.At present,the diagnosis of depression mainly depends on the clinical judgment of the doctor,the auxiliary method is the patient's self-assessment scale and other assessment scales.The above methods are affected by doctors' clinical experience,the authenticity and accuracy of patient's self-description,they are subjective and have a relatively high rate of misdiagnosis.In recent years,researchers have been looking for an objective evaluation method and quantitative indicators in order to identify depression objectively and effectively.Among them,research of depression recognition based on facial expression behavior is a hot topic.Some scholars believe that depression patients have more positive emotional feedback,negative emotional feedback and other behavior patterns than normal people,so in some cases,they have formed unique expression behavior patterns such as reduced positive expression and increased negative expression.Therefore,this article has conducted research on how to use facial expressions to effectively identify depression.The main innovations and contributions are as follows:(1)Quantitative research on the digital characteristics of the expression behavior patterns of depressed people,then verify and analyze relevant theories in depth.Based on the facial action unit and according to their physical meaning,this paper proposes a quantitative evaluation method of typical facial expression based on the action unit.We study the proportion of time and frequency of facial expression and the dynamic change rate of facial expression in depressed patients.Through targeted experimental design and result analysis,the unique facial expression behavior pattern of depressed patients is demonstrated with digital characteristics.(2)Based on the results of(1),using the techniques of time-frequency analysis and signal processing,advanced features of the expression behavior patterns were designed to identify depression.The comparison and statistical analysis of the matching receiver samples collected in 152 experiments verified the effectiveness of the feature set,which laid the foundation for a depression recognition model based on expression behavior patterns.(3)Based on the in-depth analysis of the relationship between the expression behavior patterns and advanced features of depressed people,an effective depression recognition model is constructed.In the 456 videos of 152 samples,after 100 repeated 10-fold cross-validation,the recognition rate of the model for depression was: 73.48% for men and 68.43% for women.This article uses the emotional feedback theory of depressed patients as an entry point,then design a targeted expression collection experiment for depressed people.Based on the experimental data,according to the digital features of the facial action unit,the facial expressions behavior patterns of depressed patients were quantified and corresponding advanced features were constructed.On the basis of the above results,a depression recognition model based on expression behavior patterns was established,which improved the accuracy of identifying depression based on objective indicators.