فهرست نشریات علمی دسترسی آزاد ایران

The integration of deep learning into interactive art represents a paradigm shift in how audiences engage with aesthetic experience, moving from passive observation toward active, data-mediated co-creation. Despite growing interest, few empirical studies have operationalized this convergence within a unified system capable of real-time multimodal interaction. This study addresses a critical gap in the existing literature: the absence of an empirically evaluated, unified system capable of transforming real-time multimodal audience data into personalized artistic experience. To this end, a deep learning-based interactive art system was designed, implemented, and evaluated through a design-based, mixed-methods framework comprising five iterative phases-conceptual design, algorithm selection, software development, pilot testing, and interaction data analysis. The system simultaneously ingested EEG brain signals, gaze trajectories, voice, and text from 15–20 participants aged 18–45, processing these streams through a CLIP–StyleGAN2 pipeline to generate context-responsive visual outputs in real time. Quantitative analysis revealed elevated EEG cognitive engagement indices and sustained purposeful gaze attention across all sessions; generated artworks received high aesthetic scores for composition, chromatic diversity, and semantic coherence. Qualitative thematic analysis further confirmed that participants recognized the outputs as authentic reflections of their momentary inner states, regardless of prior artistic or technical experience. These findings collectively demonstrate that deep learning algorithms can bridge data-driven computation and lived aesthetic experience, and the proposed framework offers a replicable interdisciplinary model for future research at the intersection of digital arts, cognitive science, and adaptive human–machine interaction.