Research On The Construction Of Individual Health Portrait Based On Independent Multimedia Data

Health care is one of the areas with the most potential for artificial intelligence applications.With the continuous deepening of the integration of artificial intelligence and traditional medical industry,Internet consultation and self-health management APP have been widely used.The landing of these medical application scenarios is inseparable from the identification and description of individual health conditions,and the user portrait provides ideas for this.Analysing the health of the human body,drawing perfect health portraits,and providing doctors and individuals with medical support and assistance is a key part of providing precision medical services,and it is also the trend of smart medical treatment.On the one hand,the popularization of the Internet of Things and personal devices is possible by autonomously collecting multimedia data such as selfie photos and first videos.These autonomous multimedia data contain a large amount of information related to individual health status;on the other hand,with the Deep learning is representative of the development of machine learning methods,and related technologies for information extraction and processing based on multimedia data are maturing.In this context,this article aims to extract individual health information in autonomous multimedia data based on deep learning methods,including static internal factors related to congenital diseases and dynamic external factors related to daily liferelated behaviors.These health information will improve the convenience and richness of the existing next-day health portraits.The main content of this article is based on the above two aspects,including:(1)Use the self-timed multimedia data of the face selfie image to identify whether the face image suffers from face genetic syndrome,that is,to extract and classify the feature of the face image through the classification algorithm.This paper compares the performance of multiple classification networks through experiments,and finally selects the Inception-v3 network as the backbone network of the face classification model.The idea of using the confrontation generation network alleviates the lack of sufficient face data to make the model better The problem of training effect.And through the experimental design to verify the accuracy improvement,compared to the original data set without adding generated data,the model used in this article has an accuracy rate of about 8% on the fusion of the generated image and the original image data set Small increase.The result of face recognition can be used as a static internal cause in an individual’s health portrait,such as whether there is a possibility of suffering from genetic syndrome.(2)Using the first-view video,which is independent multimedia data,through indepth research on video understanding related algorithms,combined with the difference between the first-view and the third-view video to design a behavior recognition method suitable for the first-view.First,the unique characteristics of the first perspective and the reasons why the traditional behavior recognition method is not suitable for the first perspective behavior recognition are explained.A behavior recognition method that adapts the fusion information of the first perspective data set is proposed;The network structures included in the method are a 2D structure convolution network and a 3D structure convolution network.Finally,the proposed method is tested on the first perspective data set.According to the change in accuracy in the experimental results,the fusion scene The behavior recognition model of information was verified for feasibility,and the information obtained from the first-view video recognition was used as a dynamic external cause in individual health portraits,such as running and cycling.This article focuses on self-monitoring of health conditions and pre-diagnosis assistance for doctors,using individuals to independently collect multimedia data,realtime recording and pre-judgment of health-related information,depicting "health portraits",and combining individual living habits to predict "health trends" ",To provide doctors with diagnostic references,and better write a" health prescription."...