Scholarlink

<p>Filter banks have substantially improved the detection and classification of steady-state visual evoked potential (SSVEP). However, key parameters, including the number of sub-bands, the frequency ranges of band-pass filters, and the sub-band fusion weights, exert a strong influence on the recognition accuracy of SSVEP. To address these challenges, we developed an optimized decoding algorithm termed Sparrow Algorithm-Filter Bank and Task-Related Component Analysis (SA-FBTRCA), which integrates adaptive filter banks with Task-Related Component Analysis (TRCA). The algorithm utilized the Sparrow Search Algorithm (SSA) to optimize the parameters of each filter bank and employed TRCA for classification. We evaluated our method on two public datasets (Benchmark Dataset and CA Dataset). The results demonstrated its superior effectiveness, achieving accuracies of 91.27% and 81.60%, and information transfer rates (ITR) of 183.17 bits/min and 152.64 bits/min, respectively, thereby outperforming existing algorithms. Moreover, the parameter combinations derived from the proposed method exhibited greater efficacy for SSVEP decoding than those obtained via grid search. Of the parameters evaluated, sub-band fusion weights exerted the most pronounced effect on decoding accuracy, followed by the number of sub-bands; the upper cutoff frequency of the band-pass filters demonstrated the least influence. Our findings reveal that the optimization of sub-band fusion weights can effectively counteract the performance degradation associated with increasing the number of sub-bands, thereby offering a viable strategy for enhancing filter-bank-based decoding frameworks.&#xa0;Code is available (<a href="https://github.com/muningppx/zsy_second/tree/master">https://github.com/muningppx/zsy_second/tree/master</a>).</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

TRCA with sparrow-optimized adaptive filter banks for SSVEP recognition

  • 期刊论文

  • 发表日期：

    2025年10月27日

查看全文
  • Jiaofen Nan,
  • Siyuan Zhang,
  • Panpan Xu,
  • Kaifan Zhang,
  • Duan Li,
  • Fubao Zhu,
  • Yanting Li,
  • Yongquan Xia,
  • Yinghui Meng

摘要

Filter banks have substantially improved the detection and classification of steady-state visual evoked potential (SSVEP). However, key parameters, including the number of sub-bands, the frequency ranges of band-pass filters, and the sub-band fusion weights, exert a strong influence on the recognition accuracy of SSVEP. To address these challenges, we developed an optimized decoding algorithm termed Sparrow Algorithm-Filter Bank and Task-Related Component Analysis (SA-FBTRCA), which integrates adaptive filter banks with Task-Related Component Analysis (TRCA). The algorithm utilized the Sparrow Search Algorithm (SSA) to optimize the parameters of each filter bank and employed TRCA for classification. We evaluated our method on two public datasets (Benchmark Dataset and CA Dataset). The results demonstrated its superior effectiveness, achieving accuracies of 91.27% and 81.60%, and information transfer rates (ITR) of 183.17 bits/min and 152.64 bits/min, respectively, thereby outperforming existing algorithms. Moreover, the parameter combinations derived from the proposed method exhibited greater efficacy for SSVEP decoding than those obtained via grid search. Of the parameters evaluated, sub-band fusion weights exerted the most pronounced effect on decoding accuracy, followed by the number of sub-bands; the upper cutoff frequency of the band-pass filters demonstrated the least influence. Our findings reveal that the optimization of sub-band fusion weights can effectively counteract the performance degradation associated with increasing the number of sub-bands, thereby offering a viable strategy for enhancing filter-bank-based decoding frameworks. Code is available (https://github.com/muningppx/zsy_second/tree/master).