Application Research On Skeleton Data Depression Risk Recognition And Facial Image Character Recognition Based On Gait

Depression is a complex psychological disorder that requires consideration of multiple factors for diagnosis.The development of artificial intelligence provides a new method to assist in the diagnosis and treatment of depression,particularly in the use of gait features for depression risk detection.This method has many advantages: it does not require cooperation from the subject,gait acquisition has no impact on the subject,gait features are an objective biological characteristic that is difficult to disguise,and using depth cameras such as Kinect to collect gait features can provide more accurate data and is easy to operate.However,traditional collection methods have some blind spots in the viewpoint,which can lead to imprecise data collection.The existing skeleton topology graph built from collected three-dimensional coordinate data of joints is not optimized enough to fully express the depression-related gait features.The current deployment of gait collection devices is also widespread,making it difficult to effectively manage existing gait data,while facial information in gait video data can be used for subject annotation.Based on the background mentioned above,this paper carried out the following work:1.This paper proposes a gait collection scheme based on the characteristics of the Kinect depth camera.A gait dataset was established for a depression risk group of 43 individuals and a healthy control group of 57 individuals,with a total of 100 subjects selected for gait data collection.The dataset includes raw skeletal data and gait video data,which were preprocessed.2.This paper explores the correlation between gait and depression risk.Firstly,statistical analysis was used to investigate the gait features of the depression risk group and the control group.Recursive feature elimination and mutual information methods were used to select the important features between the two groups,resulting in the extraction of six corresponding features.The skeleton topology graph was optimized based on these features.The classification accuracy reached 71%,which is a 16%improvement over the original results.3.To address the issue of a lack of diversity in facial images in gait videos,this paper used various data augmentation methods,including geometric transformation,color space transformation,image information erasure,and image blur,to enhance facial images.Additionally,the GFPGAN method was used to improve the resolution of facial images in generative adversarial network approaches.The highest classification accuracy was achieved when the facial image blur was combined with GFPGAN super-resolution image enhancement,reaching 98.68%.4.In conjunction with the recognition of depression risk based on skeleton data and facial image recognition,this paper developed a desktop-level depression risk recognition system.The front-end page was completed using HTML,CSS,and JavaScript,while the back-end function was implemented using Python language,which included features such as subject data upload,depression risk analysis,and depression improvement suggestions,thereby enhancing the practicality of the model proposed in this paper.