Scholarlink

<p>Traditional speech recognition systems encounter challenges in accurately interpreting various speech nuances and contextual variations. To overcome the issue, the proposed system innovatively integrating the EEG with multi-modal deep learning to enhance the speech recognition accuracy. To improve the understanding of language, the EEG signal is integrated with the speech data to gather most crucial data. Then, the data is pre-processed to capture the most relevant features of speech and EEG data. Then, the Memory Augmented Neural Network (MANN) model is utilised for ensuring speech recognition accuracy. The experimental results of the proposed methods with 98.32% of accuracy, 1.05% of False Positive Rate (FPR), 97.28% of precision, 98.92% of sensitivity, and 99.32% of Specificity values in dynamic acoustic environments. The results showcase the proposed method efficiency in the precisely recognising the speech patterns. Overall, our EEG-integrated multi-modal deep learning framework offers a promising avenue for augmenting speech recognition systems, paving the way for more robust and context-aware speech understanding technologies. </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

EEG-Integrated Multi-Modal Deep Learning Framework for Enhancing Speech Recognition Accuracy

  • 期刊论文

  • 发表日期：

    2025年10月9日

查看全文
  • Ameer Hussain Abdul Rahman,
  • Surendar Natarajan,
  • K. Mohanraj

摘要

Traditional speech recognition systems encounter challenges in accurately interpreting various speech nuances and contextual variations. To overcome the issue, the proposed system innovatively integrating the EEG with multi-modal deep learning to enhance the speech recognition accuracy. To improve the understanding of language, the EEG signal is integrated with the speech data to gather most crucial data. Then, the data is pre-processed to capture the most relevant features of speech and EEG data. Then, the Memory Augmented Neural Network (MANN) model is utilised for ensuring speech recognition accuracy. The experimental results of the proposed methods with 98.32% of accuracy, 1.05% of False Positive Rate (FPR), 97.28% of precision, 98.92% of sensitivity, and 99.32% of Specificity values in dynamic acoustic environments. The results showcase the proposed method efficiency in the precisely recognising the speech patterns. Overall, our EEG-integrated multi-modal deep learning framework offers a promising avenue for augmenting speech recognition systems, paving the way for more robust and context-aware speech understanding technologies.