基于单源双模跨注意力Transformer的精神疲劳监测系统

打开文本图片集

中图分类号：TB9 文献标志码：A文章编号：1674-5124（2025）09-0183-08

Abstract： To address the challenges of low classification accuracy and high system complexity in real-world mental fatigue detection，a mental fatigue monitoring system is proposed based on functional near-infrared spectroscopy （fNIRS） and a dual-modality Transformer.First，a self-developed wearable prefrontal cerebral hemoglobin concentration sensor is used to colect signals during a simulated driving task， with subjective ratings and facial datarecorded simultaneously to label three mental states： non-fatigued， moderatelyfatigued， and severely fatigued.Then，the brain activation component and the often-overlooked respiration component areextracted from the concentration signals and fed into a novel single-source dual-modality cross-attention Transformer for clasification. Finally，an online validation experiment is conducted. The proposed system achieves an average offline accuracy of 96.59%± 1.8% ， significantly outperforming the single-modality Transformer 90.45%±0.62% . In the online experiment， the fatigue alerts are consistent with subjective ratings， and the vehicle damage rate at experiment completion reduced from 100% to 5% .The system enhances fatigue monitoring accuracy while reducing hardware complexity and is applicable to real-world scenarios such as maritime aerial patrols， civil aviation flights and human-vehicle co-driving.

Keywords：mental fatigue； functional near-infrared spectroscopy； cerebral hemoglobin concentration;respiration; cross-attention

0 引言

精神疲劳是人在持续进行认知活动时体验到的一种生理、心理不适，表现为注意力难以集中、警觉度下降、困倦、嗜睡等现象，增加事故和危险发生的可能性[1-2]。（剩余10723字）