Scholarlink

<p>Micro-expressions, subtle and rapid facial changes, offer crucial insights into human emotions and psychological states. Automated spotting of micro-expressions is challenging due to their brief duration and small amplitude. This paper introduces a novel method, FCAT (Facial Core Anchoring Triangle), to enhance the accuracy of micro-expression spotting. By constructing a geometric alignment based on facial key points, FCAT effectively mitigates the impact of head pose variations. Optical flow features from 13 regions of interest are extracted to capture subtle facial motions. Low-pass filtering and empirical mode decomposition are applied to suppress noise and enhance signal stability. Experiments on CAS (ME)<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(^2\)</EquationSource> <EquationSource Format="MATHML"><math> <mmultiscripts> <mrow /> <mrow /> <mn>2</mn> </mmultiscripts> </math></EquationSource> </InlineEquation> and SAMM-LV datasets demonstrate FCAT’s competitive performance, achieving F1-scores of 0.4405 and 0.3381, respectively, and outperforming existing methods in accuracy and robustness to head-motion variations and reasonable hyper-parameter perturbations. Code is available at <a href="https://github.com/hn-yang/FCAT">https://github.com/hn-yang/FCAT</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

Facial core anchoring triangle: enhancing micro-expression spotting through geometric alignment

  • 期刊论文

  • 发表日期：

    2026年4月24日

查看全文
  • Henian Yang,
  • Shucheng Huang,
  • Hualong Yu

摘要

Micro-expressions, subtle and rapid facial changes, offer crucial insights into human emotions and psychological states. Automated spotting of micro-expressions is challenging due to their brief duration and small amplitude. This paper introduces a novel method, FCAT (Facial Core Anchoring Triangle), to enhance the accuracy of micro-expression spotting. By constructing a geometric alignment based on facial key points, FCAT effectively mitigates the impact of head pose variations. Optical flow features from 13 regions of interest are extracted to capture subtle facial motions. Low-pass filtering and empirical mode decomposition are applied to suppress noise and enhance signal stability. Experiments on CAS (ME) \(^2\) 2 and SAMM-LV datasets demonstrate FCAT’s competitive performance, achieving F1-scores of 0.4405 and 0.3381, respectively, and outperforming existing methods in accuracy and robustness to head-motion variations and reasonable hyper-parameter perturbations. Code is available at https://github.com/hn-yang/FCAT.