LieRHRV system for remote lie detection using heart rate variability parameters

Moran Davoodi, Nitay Aspis, Yael Drori, Ido Weiser-Bitoun, Yael Yaniv

Research output: Contribution to journalArticlepeer-review

Abstract

The standard polygraph, or lie detector, is limited by its reliance on average heart rate, subjective examiner interpretation, and the need for direct subject contact. Remote photoplethysmography (rPPG) offers a promising contactless alternative, by using facial videos to extract heart rate variability (HRV). We introduce "LieRHRV," a remote lie detection algorithm based solely on extracted HRV parameters. To test the HRV parameter quality, we compared these parameters to HRV parameters extracted from ECG and photoplethysmography (PPG) records archived in five gold-standard ECG/PPG datasets. A prospective study of 39 healthy volunteers was also performed to evaluate the accuracy of lie detection based on PPG- or rPPG-derived HRV parameters. Effective HRV parameter extraction from both PPG and ECG sources was demonstrated, with comparable outcomes among 60% of the parameters on average with the publicly available datasets, and prospective study with 80% of the parameters. LieRHRV performance on ECG, PPG or rPPG (with parameters selected for PPG) exhibited an accuracy of 83.3 ± 3%, 87.3 ± 4% or 91.7 ± 3.5%, respectively. In comparison, the naïve model for ECG, PPG or rPPG data achieved an accuracy of 58.3 ± 3%, 61.0 ± 3% or 67.0 ± 5%, respectively. This study demonstrated the feasibility and effectiveness of LieRHRV, and offers a promising avenue for advancing lie detection technologies beyond polygraph limitations.

Original languageEnglish
Article number30749
JournalScientific Reports
Volume14
Issue number1
DOIs
StatePublished - Dec 2024

Keywords

  • Camera
  • Machine learning
  • Remote polygraph

All Science Journal Classification (ASJC) codes

  • General

Cite this