Scholarlink

<p>This study explored the feasibility of classifying words based on their semantic and phonetic properties in Imagined Speech using EEG. We implemented a subject-specific classification approach to determine: whether distinct word properties could be reliably decoded from neural activity, and how these linguistic features influenced classification accuracy. Our analysis focused on 12 Italian words carefully selected to enable an innovative decoding paradigm. The original outcomes of the work are obtained from three novel classification paradigms whose performance are based on an innovative distance metric. In the study we comprehensively evaluate features previously reported in the literature, with the goal of identifying the most critical features and the neural channels for each property’s classification. The results demonstrate the efficacy of decoding imagined word properties, with average accuracies of 55.8% for Length, 57.2% for Double letter presence, and 35.2% for semantic category. The spatial patterns observed, particularly the robust activation of the left temporal lobe in the Alpha and Theta band, reinforce the involvement of classical language areas in Imaginary Speech and underscore the significance of frequency-specific neural dynamics in cognitive and linguistic functions.</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

A Novel Imagined Speech Paradigm: Word Classification from Neurophysiological Signals

  • 期刊论文

  • 发表日期：

    2026年1月19日

查看全文
  • Francesco Iacomi,
  • Andrea Farabbi,
  • Maximiliano Mollura,
  • Edoardo Maria Polo,
  • Riccardo Barbieri,
  • Luca Mainardi

摘要

This study explored the feasibility of classifying words based on their semantic and phonetic properties in Imagined Speech using EEG. We implemented a subject-specific classification approach to determine: whether distinct word properties could be reliably decoded from neural activity, and how these linguistic features influenced classification accuracy. Our analysis focused on 12 Italian words carefully selected to enable an innovative decoding paradigm. The original outcomes of the work are obtained from three novel classification paradigms whose performance are based on an innovative distance metric. In the study we comprehensively evaluate features previously reported in the literature, with the goal of identifying the most critical features and the neural channels for each property’s classification. The results demonstrate the efficacy of decoding imagined word properties, with average accuracies of 55.8% for Length, 57.2% for Double letter presence, and 35.2% for semantic category. The spatial patterns observed, particularly the robust activation of the left temporal lobe in the Alpha and Theta band, reinforce the involvement of classical language areas in Imaginary Speech and underscore the significance of frequency-specific neural dynamics in cognitive and linguistic functions.