@inproceedings{59727085e3be42fc9d9e25ed43fe446e,
title = "Automatic long-term deception detection in group interaction videos",
abstract = "Most work on automated deception detection (ADD) in video has two restrictions: (i) it focuses on a video of one person, and (ii) it focuses on a single act of deception in a one or two minute video. In this paper, we propose a new ADD framework which captures long term deception in a group setting. We study deception in the well-known Resistance game (like Mafia and Werewolf) which consists of 5-8 players of whom 2-3 are spies. Spies are deceptive throughout the game (typically 30-65 minutes) to keep their identity hidden. We develop an ensemble predictive model to identify spies in Resistance videos. We show that features from low-level and high-level video analysis are insufficient, but when combined with a new class of features that we call LiarRank, produce the best results. We achieve AUCs of over 0.70 in a fully automated setting.",
keywords = "Deception detection, Media understanding, Multimodal analysis",
author = "Chongyang Bai and Maksim Bolonkin and Judee Burgoon and Chao Chen and Norah Dunbar and Bharat Singh and Subrahmanian, {V. S.} and Zhe Wu",
note = "Funding Information: We presented an ensemble based automated deception detection framework called LiarOrNot which predicts deception in a group setting by processing long videos. Our framework utilizes appropriate representations at differenttemporal resolutions for multiple features which capture low and high level information. We also propose a novel class of meta-features called LiarRank which provides a significant boost in over-all performance. We evaluated LiarOrNot on a dataset collected across different sites and cultures. In a rigorous cross-validation based testing protocol, which separates identities and games during training and inference, we obtained an AUC greater than 0.7, which was 12% better than average human performance. Role of Authors. Authors Burgoon and Dunbar designed the Resistance-style game, designed how the game would be run face to face, and collected the Resistance data. The remaining authors designed the feature extraction and machine learning algorithms and software, and designed/ran all experiments. Acknowledgement. This work was funded by ARO Grant W911NF1610342. Publisher Copyright: {\textcopyright} 2019 IEEE.; 2019 IEEE International Conference on Multimedia and Expo, ICME 2019 ; Conference date: 08-07-2019 Through 12-07-2019",
year = "2019",
month = jul,
doi = "10.1109/ICME.2019.00276",
language = "English (US)",
series = "Proceedings - IEEE International Conference on Multimedia and Expo",
publisher = "IEEE Computer Society",
pages = "1600--1605",
booktitle = "Proceedings - 2019 IEEE International Conference on Multimedia and Expo, ICME 2019",
}