An LLM-Based Interaction System with Multimodal Emotion Recognition and Self-Learning Mechanism in Intelligent Electronic Pets

Ruize Wang

2025

Abstract

This study presents an LLM-based interaction system for intelligent electronic pets, aiming to enhance emotional adaptability and personalized feedback through multimodal perception technology and self-learning. Traditional electronic pets rely on fixed interaction modes, limiting their ability to provide individualized emotional responses. The proposed system, structured into three layers—perception, decision, and execution—uses various sensors to gather user emotions, environmental data, and preferences, then applies LLM technologies like GPT-4 to generate adaptive feedback. The system's self-learning capability continuously optimizes responses based on evolving user interactions. Virtual user samples were created to simulate system decision-making, and the feedback was evaluated across multiple dimensions. Results showed superior emotional alignment, feedback diversity, and adaptability compared to unimodal and rule-based models, highlighting the system's exceptional self-learning capabilities. This research underscores the critical role of LLMs in multimodal emotion processing and self-learning, offering theoretical and technical guidance for the use of intelligent electronic pets in emotional support and companionship applications.

Paper Citation

in Harvard Style

Wang R. (2025). An LLM-Based Interaction System with Multimodal Emotion Recognition and Self-Learning Mechanism in Intelligent Electronic Pets. In Proceedings of the 2nd International Conference on Engineering Management, Information Technology and Intelligence - Volume 1: EMITI; ISBN 978-989-758-792-4, SciTePress, pages 423-431. DOI: 10.5220/0014360700004718

in Bibtex Style

@conference{emiti25,
author={Ruize Wang},
title={An LLM-Based Interaction System with Multimodal Emotion Recognition and Self-Learning Mechanism in Intelligent Electronic Pets},
booktitle={Proceedings of the 2nd International Conference on Engineering Management, Information Technology and Intelligence - Volume 1: EMITI},
year={2025},
pages={423-431},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0014360700004718},
isbn={978-989-758-792-4},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 2nd International Conference on Engineering Management, Information Technology and Intelligence - Volume 1: EMITI
TI - An LLM-Based Interaction System with Multimodal Emotion Recognition and Self-Learning Mechanism in Intelligent Electronic Pets
SN - 978-989-758-792-4
AU - Wang R.
PY - 2025
SP - 423
EP - 431
DO - 10.5220/0014360700004718
PB - SciTePress