Scholarlink

<p>Human brain seamlessly integrates multisensory stimuli to synthesize complementary information for enhanced perceptions, depending on neural principles of superadditivity, inverse effectiveness, and temporal congruency. Replicating multisensory integration in artificial intelligences has remained challenging due to the inefficiency of algorithmic fusions and the absence of hardware-native mechanisms. Here, we demonstrate biomimetic audiovisual integration at the device level of Bi<sub>2</sub>O<sub>2</sub>Se ferroelectric-semiconductor field-effect transistors (FeS-FETs) through multiphysics coupling. Our FeS-FETs simultaneously accomplish the superadditive integration factor of 2800%, dynamical reweighting inputs of inverse effectiveness, and prolonged temporal congruency beyond 10<sup>3</sup> s. Furthermore, when configured into memristor-chip-based spiking neural networks, the resultant multisensory system is capable of executing the sensory synaptic plasticity, population-coded spiking, and Bayesian-optimal fusion, which promotes the excellent recognition accuracy of 98.2% for fuzzy objects, surpassing that identified from conventional fusion algorithms. By the exploration of multi-physical computing to mirror the biological multisensory hierarchy, we establish a physics-aware framework for neuromorphic multisensory intelligences, bridging physical dynamics with neurobiological principles for self-adaptive edge computing.</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

Biomimetic ferroelectric-semiconductor transistor enables neuronal multisensory integration

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月20日

查看全文
  • Shuo Liu,
  • Ligong Zhang,
  • Ruiqing Xie,
  • Zhiyuan Wu,
  • Aojie Chen,
  • Lixia Han,
  • Linbo Shan,
  • Zuoyuan Dong,
  • Junling Liu,
  • Xinrui Guo,
  • Yu Zhu,
  • Zhenghua Zhou,
  • Fei Liu,
  • Linxiao Shen,
  • Peng Huang,
  • Xinran Wang,
  • Zongwei Wang,
  • Yimao Cai,
  • Ming He

摘要

Human brain seamlessly integrates multisensory stimuli to synthesize complementary information for enhanced perceptions, depending on neural principles of superadditivity, inverse effectiveness, and temporal congruency. Replicating multisensory integration in artificial intelligences has remained challenging due to the inefficiency of algorithmic fusions and the absence of hardware-native mechanisms. Here, we demonstrate biomimetic audiovisual integration at the device level of Bi2O2Se ferroelectric-semiconductor field-effect transistors (FeS-FETs) through multiphysics coupling. Our FeS-FETs simultaneously accomplish the superadditive integration factor of 2800%, dynamical reweighting inputs of inverse effectiveness, and prolonged temporal congruency beyond 103 s. Furthermore, when configured into memristor-chip-based spiking neural networks, the resultant multisensory system is capable of executing the sensory synaptic plasticity, population-coded spiking, and Bayesian-optimal fusion, which promotes the excellent recognition accuracy of 98.2% for fuzzy objects, surpassing that identified from conventional fusion algorithms. By the exploration of multi-physical computing to mirror the biological multisensory hierarchy, we establish a physics-aware framework for neuromorphic multisensory intelligences, bridging physical dynamics with neurobiological principles for self-adaptive edge computing.