Abstract

In this paper, we investigate the geometric model database of jewelry design based on the perceptual-engineering system, as jewelry products often focus on the exquisite and unique shape, the preciousness of the selected materials and the excellence of the production process. Firstly, the paper introduces the relevant analysis methods, namely structural equation analysis, Bi-LSTM and conditional random field (CRF), to be used for perceptual engineering analysis, followed by the construction of semantic imagery space. This paper implements the extraction of keywords for perceptual information about the appearance and users of jewelry design works based on a bidirectional long and short term memory neural network (BI-LSTM) and conditional random field. The analysis shows that "rough - elegant", "robust - soft", "inexpensive - high-class", "complex - simple "The mean values of the five groups of perceptual words "popular-personalized" in the 10 samples were 3.6421, 3.9451, 4.1328, 4.03791, and 3.84617, respectively, with an overall bias toward elegant, soft, high-class, simple and personalized, and a mean score of 4.1371. Finally, the establishment of jewelry design geometric model database based on the user perceptual information extraction system and the constructed semantic space is of greater significance for jewelry designers to improve their design efficiency.