While looking at glasses, we will have various sensation on their forms, colours, textures etc. These feelings direct us to decide which pair of glasses to buy. This study tries to find out the relationship between human sensation and glasses design; that is, to figure out what kinds of forms will excite what kind of human sensation. By adopting this relationship, we also propose a computer support system which can automatically generate appropriate glasses forms in response to the expectation of consumers. In this study, we first collect several adjectives which can express the human sensation on glasses designs. Then, in an experiment, subjects are asked to evaluate a set of glasses in order to acquire the ratings of those adjectives. We further analyze the formal elements included in different glasses, multivariate analysis and the neural networks are used to decide which design of each element are more likely to excite a specific human sensation. These results then are built as a knowledge base of a support system which includes an inference engine to assist a designer to produce new glasses design. By receiving the input of desired adjectives, this system will find out the most preferred elements for each adjective and integrate them into some appropriate design prototypes for further modification.