Research On Multisensory Design Of Electric Shavers From The Perspectives Of Kansei Engineering And Artificial Neural Networks

With the upgrading of the quality of consumption in the market,in addition to functional requirements,consumers are increasingly focused on the emotional imagery conveyed by the products.The single shape of the product is no longer able to meet the emotional needs of the user in a multisensory way.How products can meet consumers’ emotional needs through multisensory channels has become the focus of design research.As a product that is used frequently,electric shavers can convey rich emotional imagery in terms of shape and sound.Therefore,this study takes an electric shaver as an example,and uses Quantitative Theory Type I(QTTI)and Genetic Algorithm(GA)combined with BP Neural Network(BPNN)to quantify the emotional needs of users through sensory engineering.To explore the construction of a correlation model between design elements of shape and sound and Kansei evaluation,providing designers with clear design indicators and references in multisensory design,making the original process of design by experience more logical and scientific.Firstly,a sample of 230 electric shavers was collected through online and offline channels,and 34 representative samples were selected and clustered using focus groups,multivariate scaling and cluster analysis methods.Consumer Kansei evaluation words(135 groups)was also collected through literature research and other means,and then four groups of representative Kansei words expressing shape and sound imagery were screened by the semantic difference method.Secondly,the product design elements were classified into items and categories through morphological analysis method and audio software analysis,and a total of 7+30 item categories were obtained,so that the design elements included both shape and sound.A linear prediction model between product design elements and consumer perceptions was then constructed using QTTI.A qualitative multiple linear regression equation was constructed to derive the relevant item category scores and the degree of influence of imagery.Two different non-linear prediction models are constructed using BP neural networks and GA-BPNN to understand the difference between linear and non-linear models.Finally,paired-sample t-tests were conducted on the QTTI linear model,BPNN and GA-BPNN prediction models,and all three p-values were greater than 0.05,indicating the reliability of the linear and non-linear prediction models in this study.The mean error comparison method was also used for comparative merit selection,and the results showed that the QTTI model had better prediction accuracy in the multisensory design.Finally,the design metrics provided by the QTTI model were used to perform AI design using a text-image model,which was validated by a Likert seventh-order scale to complete the product design practice.In conclusion,this research method provides a clear reference for design metrics and predictive measures of perceptual imagery in multisensory product design and application,allowing for a more objective and logical design process that would otherwise be based on experience.