사운드 스펙트로그램과 심층신경망 기법을 사용한 패셔너블한 웨어러블 제품 디자인을 위한 테크놀로지

As digital culture became popular, various cultural styles and customs were shared. This has created a tendency to create new genre styles. As a result of this phenomenon, The development of the design which fuses fashion and technology is widely spread. Wearable products in new genres have been introduced into the market with much expectation. However, market of wearable products is in chaos. The introduction of products with an emphasis on technical functions has been analyzed as one of the reasons for the sluggishness. One of the ways to revitalize the market, due to the characteristic of wearable products, worn as an outer wear, amalgamation of the fashion element, which expresses an identity of a user, and the unique value of design that enables the automatic reaction of artificial intelligence should be implemented. This paper proposes a technological system with a transformation map from sound to graphic image that can be used for designing auto-responsive and fashionable wearable products that amalgamate fashion and artificial intelligence in the stylishly designed display. The display will automatically react to environmental changes and show predefined fashionable display designs without users’ awareness. After the sound spectrogram extracts the characteristics of sound input from the surrounding environment to classify and label them through Keras method of DNN, it automatically reacts to covert the sound input into stylish graphic images and display them fashionably according to the user’s preset design. The result of the experiment with the proposed sound system has proved that the system is capable of recognizing and characterizing the sound inputs around the wearable products when the noise level of the environment is under –10 dB. The experiment has verified that inputted sound can be transformed into a graphic image according to classified labels and it is successfully put on display.