{
  "id": 542670,
  "title": "Unraveling emotional signatures: comparing physiological methods and algorithm-based recognition of spontaneous emotional facial expressions",
  "url": "https://urgent.news/2026/08/10/unraveling-emotional-signatures-comparing-physiological-methods-and",
  "topic": "science",
  "section": "Science",
  "published": "2026-08-10T00:00:00.000Z",
  "source": {
    "name": "bioRxiv",
    "slug": "biorxiv",
    "url": "https://www.biorxiv.org/content/10.64898/2026.08.05.742947v1?rss=1"
  },
  "original_language": "en",
  "account": "Understanding and interpreting the emotions of others is crucial for social interactions. While traditional biological psychology relies on laboratory measures to infer emotional responses, contemporary computer vision algorithms attempt to detect emotions from facial video without explicit communication intent. However, the effectiveness of these algorithms in classifying spontaneous emotional reactions without purposeful communication is uncertain. To investigate, we compared established psychophysiological measures (EEG, facial EMG, EDA activity) with the open-source facial behavior toolkit OpenFace for classifying participants' spontaneous reactions while viewing happiness-inducing, disgust-inducing, and neutral pictures. Participants rated the emotional valence and arousal, and then selected the basic emotion that best matched their reaction, serving as the classification criterion.\n\nEmploying within-participants single-trial support vector machine (SVM) classification, EEG demonstrated the highest accuracy at 40%, followed by facial EMG at 37%; OpenFace achieved 36%. All methods, except EDA, surpassed chance performance of 33.3% and were lower compared to human raters at 48%. Performance for across-participants SVMs declined slightly, hitting chance level for OpenFace and EDA. The study suggests that both psychophysiological measures and video-derived facial action units can capture diagnostic aspects of emotional responding during picture viewing, but their performance is limited when expressions are spontaneous and uncommunicative. Individual differences in expressivity and physiological responses contribute to these limitations and should be considered when applying automatic emotion recognition in research or practical scenarios.",
  "summary": "Recognizing others emotions is central to social interaction. Traditional biological psychology infers emotional responding via laboratory measures, whereas contemporary computer vision algorithms claim to identify emotions unobtrusively from facial video. However, the validity of such algorithms for classifying spontaneous emotional responses occurring without explicit communicative intent…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}