Optimisation and Application of Self-Supervised Transformer Model Based on Particle Swarm Optimisation in Emotion-Centred Product Design: - A Case Study of Beauty Appliances

As a popular beauty product in recent years, beauty instruments have gained significant attention, with their appearance design showcasing diverse stylistic features. However, limited emphasis has been placed on enhancing the user's emotional experience during product use. This paper applies theories of perceptual engineering to the aesthetic design of beauty instruments, with a focus on analyzing and improving users' emotional interactions. Multimodal information, including users' text feedback and facial expression da-ta, is utilized for emotion analysis. To enhance the accuracy of this analysis, the Transformer model is employed. Considering that the manual adjustment of hyperparameters (e.g., learning rate, attention head count, hidden layer dimensions) is time-intensive and inefficient, a particle swarm optimization algorithm is introduced to automate the hyperparameter tuning process, thereby improving the performance of emotion analysis.