Human-Computer Interaction (SIG HCI)
Loading...
Paper Type
Complete
Paper Number
1308
Description
This study aims to investigate whether nonverbal features before an act of deceit entail discriminative ability for deception detection and to find the optimal time duration for data collection and analysis when designing multimodal systems for deception detection. An automated interviewing system was proposed to collect and analyze multimodal behavioral data from participants during rapid screening with a fraudulent document scenario. Features for response latency and facial action units were extracted from the silent moment after the end of a question and prior to participants’ vocal responses. Facial action units and vocalic features were also extracted during the vocal responses. Prediction results showed that silence-time features improved deception detection accuracy from random guesses. Additionally, using silence-time and speech-time features jointly outperformed using silence-time features but not speech-time features.
Recommended Citation
Wang, Xinran; Ge, Saiying; Chen, Xunyu; and Walls, Bradley L., "Deception Detection with Nonverbal Behaviors from Silence and Speech Time" (2021). AMCIS 2021 Proceedings. 8.
https://aisel.aisnet.org/amcis2021/sig_hci/sig_hci/8
Deception Detection with Nonverbal Behaviors from Silence and Speech Time
This study aims to investigate whether nonverbal features before an act of deceit entail discriminative ability for deception detection and to find the optimal time duration for data collection and analysis when designing multimodal systems for deception detection. An automated interviewing system was proposed to collect and analyze multimodal behavioral data from participants during rapid screening with a fraudulent document scenario. Features for response latency and facial action units were extracted from the silent moment after the end of a question and prior to participants’ vocal responses. Facial action units and vocalic features were also extracted during the vocal responses. Prediction results showed that silence-time features improved deception detection accuracy from random guesses. Additionally, using silence-time and speech-time features jointly outperformed using silence-time features but not speech-time features.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.