Research On Product Design Evaluation Methods Based On Multi-sensory

The paper gives an account of study on product design evaluation method by means of multi-sensory analyses. Main work include the following aspects: data collection and pre-process of sensory description, reduction of multi-sensory evaluation indicators, intelligent assessment of sensory attributes, eye tracking technique in evaluating product design, and multi-sensory based evaluation integrated method.In the study of data collection and pre-process of sensory descriptions, the diversification for sensory sources where evaluating messages originally come from is proposed, and a primary multi-sensory evaluation model is set up in the case of the design of community public information service facilities. Following the analyses of information characteristics of evaluation indicators and the construction of a data pretreatment model, the uniformization pretreatment of primary evaluation information is achieved in terms of the form and granularity.In the study of reduction of multi-sensory evaluation indicator, an indicator reduction model is set up based on 2-tuple rough set method which also includes a new discretization method for 2-tuple sensory description as well as discretization rules that are based on 2-tuple number. By using the indicator reduction model, the primary multi-sensory indicators in the case of publish information service facilities design are processed with efficiency, based on which the model comprising indicators for multi-sensory evaluation of public information service facilities is established.In the study of intelligent assessment of sensory attributes, the paper compares the prediction results as well as factors affecting the accuracy of the three hybrid neural network models including GA-BP model(Genetic Algorithm Optimized Back Propagation), GY-BP model(Grey Series-Back Propagation) and GY-GA-BP model(Grey Series-Genetic Algorithm Optimized Back Propagation). It also analyses the deficiency of GM(Grey Static Model) by comparing it with BP models. As a result, the paper comes to the conclusion that a flexible use of hybrid neural network tools is necessary for sensory evaluation. By comparing the prediction results of GY-GA-BP using two kinds of judement data-before rduction and after reduction, it indicates that the 2-tuple and rough set based reduction model is able to effectively reduce the number of indicators, thus increasing the stability of the prediction model.In the respect of introducing eye tracking technique in product design evaluation, a set of experiments are carried out using RED(Red-Eye Detection) system produced by SMI company, and 38 groups of valid eye movement test data and 45 copies of questionnaire survey data are collected. Through multi-factor ANOVA(Analysis of Variance), independent-samples T test and One-Way ANOVA on SPSS for Windows 16.0 of five KPIs(Key Point Index) of eye tracker including KPI2(entry time), KPI3(dwell time), KPI6(regression number), KPI8(first fixation duration) and KPI9(fixation point number), the results in different treatments as affected by group, environment and program design factors are laid out. Meanwhile, visual attribute data from the questionnaire survey are analyzed by GY-GA-BP neural network model. By comparing the test results and intelligent evaluation results, the design of eye movement experiment and the adoption of GY-GA-BP neural network model in this regard are verified.In the study of multi-sensory evaluation integrated method, a L-AHP(Linguistic-Analytic Hierarchy Process) weight definition method based on linguistic assessment, which improves upon the scaling system of pairwise comparison of the conventional AHP weight definition method, is presented, a weight conversion equation for linguistic assessment sets with varied granularity is given and an integrated TOPSIS(Technique for Order Preference by Similarity to Ideal Solution) product design evaluation model based on multi-sensory description is constructed. Combining GY-GA-BP model and above methods, an integrated evaluation based on multi-sensory description concerning the design of community public information service facilities is achieved which at the same time is validated by questionnaire survey. The methods are proved to be effective.