Design of Large Model-Based Emotional Companion Human-Computer Interaction System

With the rapid development of artificial intelligence, emotional companion human-machine interaction systems have become a research hotspot, but existing systems lack in-depth emotional understanding and natural response generation, so this study aims to build a large model-based system to enhance emotional interaction quality.Existing methods relying on rules or simple machine learning struggle to handle complex emotions and generate rigid responses; our method integrates multimodal emotion recognition with large models, using prompt engineering and optimization to improve understanding accuracy and response naturalness; this bridges technical and emotional interaction gaps for more human-like experiences.Experimental results show the system achieves 85% emotion recognition accuracy (42% higher than traditional systems) and 4.2/5 in emotional congruence, demonstrating its value in advancing emotional human-machine interaction.