Scholarlink

<p>Brain-derived emotion recognition is an explosively developing area in the field of affective computing, particularly within the context of real-time brain-computer interaction (BCI), and emotional modulation of music. Real-time accurate interpretation of data is an extreme problem because of the unpredictable and high-dimensional nature of brain activity despite the fact that EEG signals offer a good and non-invasive means of monitoring a dynamic emotional response. To realize real-time extraction of emotional state of emotional experiences (emotional valence or arousal) from EEG sensors that were induced by musical stimulation, we introduce a novel hybrid deep learning architecture that combines Temporal Convolutional Networks (TCNs) and Capsule Networks (CapsNets). Eight healthy volunteers participated; their EEG data were collected during the three sessions and eighteen trials in each of the sessions. All of the music clips were designed to evoke certain emotional changes. TCN-CapsNet model exploits the temporal dynamics of the EEG data by utilizing dilated Convulsions and dynamic routing in capsules in a way to benefit from the spatial relationships. Its highest accuracy is 89.14% and inter-session and inter-subject generalizability is high ensuring that the experiments on its model perform better than the traditional methods on three benchmark datasets. Usability and Retrieval Evaluation with Models of Information further enhances the system’s ability to adapt to diverse participants, providing insights into how well the model retrieves and utilizes emotional data to enhance user experience. This work combines a trustworthy, real-time emotion recognition system to carry out adaptive musical interaction by usage of EEG-based BCIs pertaining to emotion.</p>

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

Music emotion interaction system based on brain-computer interface: a hybrid deep learning and transfer learning framework

  • 期刊论文

  • 发表日期：

    2026年1月2日

查看全文
  • Cui yanbin

摘要

Brain-derived emotion recognition is an explosively developing area in the field of affective computing, particularly within the context of real-time brain-computer interaction (BCI), and emotional modulation of music. Real-time accurate interpretation of data is an extreme problem because of the unpredictable and high-dimensional nature of brain activity despite the fact that EEG signals offer a good and non-invasive means of monitoring a dynamic emotional response. To realize real-time extraction of emotional state of emotional experiences (emotional valence or arousal) from EEG sensors that were induced by musical stimulation, we introduce a novel hybrid deep learning architecture that combines Temporal Convolutional Networks (TCNs) and Capsule Networks (CapsNets). Eight healthy volunteers participated; their EEG data were collected during the three sessions and eighteen trials in each of the sessions. All of the music clips were designed to evoke certain emotional changes. TCN-CapsNet model exploits the temporal dynamics of the EEG data by utilizing dilated Convulsions and dynamic routing in capsules in a way to benefit from the spatial relationships. Its highest accuracy is 89.14% and inter-session and inter-subject generalizability is high ensuring that the experiments on its model perform better than the traditional methods on three benchmark datasets. Usability and Retrieval Evaluation with Models of Information further enhances the system’s ability to adapt to diverse participants, providing insights into how well the model retrieves and utilizes emotional data to enhance user experience. This work combines a trustworthy, real-time emotion recognition system to carry out adaptive musical interaction by usage of EEG-based BCIs pertaining to emotion.