Research On Virtual Nursing Home System Based On Hierarchical Computing And Artificial Intelligence

The increasing aging of the population poses greater challenges to the national pension system and services,and most of the elderly in my country still choose the traditional home-based pension model.This paper proposes a virtual nursing home research based on artificial intelligence and hierarchical computing,using a variety of technologies in the field of artificial intelligence to collect information about the elderly at home and obtain service requests from the elderly.In the event of an accident in the elderly,the relatives should be notified,or rescue measures should be taken directly.As a new development branch of the national pension mechanism,virtual nursing homes are a key step for the country's development to a new height.The addition of AI technology makes the system compared to traditional nursing home systems.The virtual nursing home economy uses hierarchical computing theory and new artificial intelligence technologies.The main work of this paper includes:1.A system analysis method based on the hierarchical computing theory,which divides a complex system such as a virtual nursing home using the hierarchical computing theory.The introduction of granular space can divide different levels and layer different parts of the system.Processing,the separate design of system software and hardware,the refinement of the various functional modules of the system,are divided into different granularities,and then combined to design a fully functional virtual nursing home system.2.Research on the optimization of human pose estimation based on the OpenPose model framework.In order to solve the technical problems of virtual nursing homes,under the original OpenPose network model,the network is light-weighted,and the lightweight network Mobile Net V3 is used to replace the VGG network,reducing the amount of network parameters,reducing the amount of network calculations,and improving network calculations speed.In order to ensure the receptive field state,a hollow convolution is added,and a convolution block with a small convolution kernel is used to replace some large convolution kernels to improve accuracy.After testing and training,based on the original model,the accuracy of this model is increased by 6.9%,which is an increase of 2.9% compared to Light?OpenPose under the premise of ensuring speed.Single-person estimation is more accurate for more people's pose estimation.Considering that the target of this article is the elderly who live alone or whose relatives are not nearby temporarily,single-person pose estimation is more used,and the error in the multi-person scenario does not affect the research of the subject.In addition,the system adds a Kinect body sensor to replace the ordinary camera,which further improves the accuracy and more closely simulates the real scene.Based on the hierarchical computing and artificial intelligence system design,through simulation experiments,the system's request response and fall detection functions are tested,which proves the researchability of this method and the future development space.