Research On Evaluation Of Elderly-Oriented Product Modeling And Images Based On ResNet Network

With the development of the society,China has entered the aging society.As China has higher and higher population aging degrees,its seniors markets also expand accordingly.The elderly have their own characteristics in psychology,physiology and other aspects,so they have diversified emotional demands for products.Because of the ambiguous and abstract user experience of products,we still lack enough considerations on physiological,psychological and cognitive demands of the elderly.Besides,researches of related fields are mainly done by questionnaire survey,user survey and other qualitative research methods.Some designers often design products for the elderly according to their experience and inspiration,but they lack the support of detailed,scientific and reasonable data.People rarely explore elderly-oriented products with the aim to quantify user experience of seniors.Besides,there is still a lack of effective methods which can correctly identify and grasp the relationship between elderly-oriented products and images and the design of elderly-oriented products.Therefore,it is relatively difficult to transform emotional demands of the elderly into the design of elderly-oriented products.This paper puts forward an evaluation model for elderly-oriented product modeling and images based on Neural Networks and Deep Learning and its concrete work includes the following four parts:Part 1 introduces current situations of researches and application of elderly-oriented products,Kansei Engineering and artificial neural networks and puts forward research method and framework of this paper.Part 2 uses the questionnaire method and the in-depth interview method to investigate life of elderly users so as to explore elderly-oriented sentimental demands.Besides,typical old user profiles are also built to know old people’s usage of daily life products and their expectations for new products and semantic differential method and clustering methodology are used to obtain a lexical library for elderly-oriented product modeling and images which include 60 groups of vocabulary sets.After investigating and surveying life scenes of the elderly,this paper determines TV remote control as the research sample for evaluation models of elderly-oriented product modeling and images.In order to establish an artificial neural network model to evaluate elderly-oriented product modeling and images,Part 3 configures network development environment first of all.Then,Part 3 crawls thousands of pieces of TV remote control images and data in the way of web crawler so as to expand the sample data set and increase the sample size and carries out multi-label semantic annotations to TV remote control images and data.Finally,this paper designs a Deep Residual Network(ResNet)learning model with the TV remote control images and data as the input neuron.The computer programming technology is used to simulate and predict people’s nonlinear thinking of perceptual image evaluation of products and high prediction accuracy is obtained through multiple iterations and learning.Part 4 inputs the new design scheme for TV remote control to the evaluation model for rating so as to evaluate the effectiveness of neural network model for evaluation of elderly-oriented product modeling and images.Part 4 also compares satisfaction degrees of elderly users and designers so as to verify the feasibility of the model.The neural network learning model is used to build an evaluation model for elderly-oriented products and images,which can give guidance to evaluation process of modeling design of elderly-oriented products,can improve design efficiency of elderly-oriented products and reduce the time cost.