Scholarlink

Emotion is an indispensable part of daily interpersonal communication, and people can communicate through emotional expression. However, the social trend of younger people suffering from psychological problems has attracted widespread attention. Aiming at children’s mental health, this paper discusses a proposed dual-path attention-enhanced graph convolutional network for children’s speech emotion recognition, and also using the attention mechanism to enhance the model’s ability to understand complex speech signals, learn the changes in emotional time information, and classify children’s speech emotions. The performance of the proposed Hierarchical Dual-Attention Network classifier is extensively evaluated. The results show that compared with existing models provided by other published studies, it achieves significant recognition improvements, recognizing 86.49% in the integrated dataset with the FAU-Aibo dataset, C-BESD dataset, and MESD dataset. This study provides new ideas and technical support for children’s speech emotion recognition and has potential application value in children’s mental health assessment, intelligent education, and human-computer interaction.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Dual-Path Attention-Enhanced Graph Convolutional Network for Children’s Speech Emotion Recognition

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Deneng Tang,
  • Liwei Lin,
  • Haojie Huang,
  • Qixuan Liu

摘要

Emotion is an indispensable part of daily interpersonal communication, and people can communicate through emotional expression. However, the social trend of younger people suffering from psychological problems has attracted widespread attention. Aiming at children’s mental health, this paper discusses a proposed dual-path attention-enhanced graph convolutional network for children’s speech emotion recognition, and also using the attention mechanism to enhance the model’s ability to understand complex speech signals, learn the changes in emotional time information, and classify children’s speech emotions. The performance of the proposed Hierarchical Dual-Attention Network classifier is extensively evaluated. The results show that compared with existing models provided by other published studies, it achieves significant recognition improvements, recognizing 86.49% in the integrated dataset with the FAU-Aibo dataset, C-BESD dataset, and MESD dataset. This study provides new ideas and technical support for children’s speech emotion recognition and has potential application value in children’s mental health assessment, intelligent education, and human-computer interaction.