By analyzing the consumer's perceptual demand and the design elements of the multifunctional tea table product form, the paper aims to establish the regression relationship model between the two to complete the personalized design of the multifunctional tea table product and solve the problem that the tea table product cannot be designed and manufactured according to the user's consumption demand. Firstly, semantic differential method is used for consumer products for tea table perceptual image value, and the factor analysis is used to summarize the value of sorting. Secondly, according to the product design elements on the tea table module deconstruction, and each part of the module is numerically coded. And BP neural network of tea table products is trained according to the perceptual image evaluation value and module value, and the mapping relationship between them is established. Finally, the accuracy of the BP neural network is verified by the questionnaire experiment with the second semantic difference method. According to the BP neural network of tea table products trained, the model of tea table products with the highest perceptual evaluation value can be predicted. The accuracy of BP neural network model of tea table products is verified by the experimental results of the second semantic difference method, which provides favorable support for the personalized design of tea table products. This method improves the design efficiency and rationality of coffee table products, solves the problem that furniture designers cannot accurately complete objective product design according to users' subjective needs, and provides beneficial reference and guidance for product design and manufacturing based on consumer demand and market.