A Multimodal Emotion Recognition Framework for Dynamic Content Adaptation and Market Response Prediction in Mobile Advertising

Abstract

With the proliferation of mobile internet and the widespread use of smart devices, mobile advertising has emerged as a central channel for enterprise promotion. However, challenges such as content homogeneity and insufficient alignment with users’ real-time emotional states have constrained advertising efficiency and diminished user experience. Accurately identifying user emotions and dynamically adapting advertising content to enhance market responsiveness remains an urgent research imperative. Existing studies on emotion recognition and content optimization in mobile advertising exhibit several limitations: many rely solely on unimodal emotion recognition, neglecting the complementary nature of multimodal signals and thereby limiting recognition accuracy; content adaptation approaches are predominantly static, lacking mechanisms for real-time dynamic adjustment; and market response prediction models often fail to integrate multimodal emotional features, resulting in suboptimal prediction accuracy and generalizability. To address these gaps, a multimodal emotion recognition model for mobile advertising was developed, accompanied by a dynamic content adaptation system driven by fused multimodal emotional features. Furthermore, a market response prediction mechanism grounded in multimodal emotional feature fusion was established. These contributions fill critical theoretical gaps in the areas of static content adjustment and unimodal prediction and significantly enrich the theoretical framework for dynamic optimization in mobile advertising.