Eye-blink is a sensitive index of cognitive load and some studies have reported that it can be a useful cue for detecting deception. However, it is difficult to apply in the real forensic scenario as very complex techniques to record eye blinking are usually needed (e.g., electrooculography, eye tracker technology). In this paper, we propose a new approach to automatically detect eye blinking based on a computer vision algorithm, which does not require any expensive technology to record data. Results demonstrated that the automatic blink detector reached an accuracy similar to the electrooculogram in detecting the blink rate. Moreover, the automatic blink detector was applied to 68 videos of people who were lying or telling the truth about a past holiday, testing the difference between the two groups in terms of blink rate and response timing. Training machine learning classification models on these features, an accuracy up to 70% in identifying liars and truth-tellers was obtained.

Using blink rate to detect deception: A study to validate an automatic blink detector and a new dataset of videos from liars and truth-tellers

Curci A.;
2020-01-01

Abstract

Eye-blink is a sensitive index of cognitive load and some studies have reported that it can be a useful cue for detecting deception. However, it is difficult to apply in the real forensic scenario as very complex techniques to record eye blinking are usually needed (e.g., electrooculography, eye tracker technology). In this paper, we propose a new approach to automatically detect eye blinking based on a computer vision algorithm, which does not require any expensive technology to record data. Results demonstrated that the automatic blink detector reached an accuracy similar to the electrooculogram in detecting the blink rate. Moreover, the automatic blink detector was applied to 68 videos of people who were lying or telling the truth about a past holiday, testing the difference between the two groups in terms of blink rate and response timing. Training machine learning classification models on these features, an accuracy up to 70% in identifying liars and truth-tellers was obtained.
2020
978-3-030-49064-5
978-3-030-49065-2
File in questo prodotto:
File Dimensione Formato  
Monaro et al., 2020.pdf

non disponibili

Tipologia: Documento in Versione Editoriale
Licenza: NON PUBBLICO - Accesso privato/ristretto
Dimensione 735.96 kB
Formato Adobe PDF
735.96 kB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11586/363142
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 3
  • ???jsp.display-item.citation.isi??? ND
social impact