The Analysis of Body Emotion Recognition in New Media Art Exhibition Space-Based Artificial Intelligence

This study aims to explore the relationship between Artificial Intelligence (AI)-driven new media art exhibition space design and immersive experiences, incorporating body-based emotional recognition methods to optimize and innovate the design of exhibition spaces from an emotional perspective. The objectives include investigating the feasibility and effectiveness of body-based emotional recognition methods in art exhibition space design and developing an AI-based new media art exhibition space to enhance viewers' immersive experiences. The validity and practicality of the proposed methods are verified through experiments and evaluations. The results demonstrate that: (1) The proposed method excels in emotional recognition tasks. On the Emotions in Context (EMOTIC) dataset, the accuracy of the proposed method is 0.897, and the F1-Score is 0.887, significantly outperforming the Vision Transformer (ViT) and Swin Transformer models, which have accuracy scores of 0.872 and 0.889, respectively. Precision and recall are also impressive, at 0.890 and 0.885, respectively. On the GroupWalk dataset, the proposed method achieves an accuracy of 0.885 and an F1-Score of 0.877, whereas the ViT and Swin Transformer models have accuracy scores of 0.857 and 0.874, respectively. These results highlight the efficiency and robustness of the proposed method in handling body language emotional recognition. (2) Through practical application testing in immersive exhibition spaces, the average recognition rates for pleasure, immersion, and awe are observed to be 0.900, 0.855, and 0.890, respectively. The average recognition rates for disgust, disappointment, surprise, and anticipation are 0.082, 0.075, 0.875, and 0.842, respectively. These data indicate that immersive exhibition space design effectively enhances viewers' feelings of pleasure, immersion, and awe while reducing feelings of disgust and disappointment. These findings validate the effectiveness and superiority of the proposed method in emotional recognition tasks and provide significant insights for further research and application of emotional recognition technology.