Abstract: Humans ability to detect lies is no more accurate than chance according to the American Psychological Association. The state-of-the-art deception detection methods, such as deception detection stem from early theories and polygraph have proven to be unreliable. Recent advancement in deception detection includes the application of advanced data analysis and machine learning algorithms. This paper presents a novel deep learning driven multimodal fusion for automated deception detection, incorporating audio cues for the first time along with the visual and textual cues. The critical analysis and comparison of the proposed deep convolutional neural network (CNN) based approach with the state-of-the-art multimodal fusion methods have revealed significant performance improvement up to 96% as compared to the 82% prediction accuracy reported in the recent literature.

Pdf: http://mandargogate.github.io/papers/SSCI2017-Deep-Learning-Multimodal-Deception.pdf


  author    = {Mandar Gogate and
               Ahsan Adeel and
               Amir Hussain},
  title     = {Deep learning driven multimodal fusion for automated deception detection},
  booktitle = {2017 {IEEE} Symposium Series on Computational Intelligence, {SSCI}
               2017, Honolulu, HI, USA, November 27 - Dec. 1, 2017},
  pages     = {1--6},
  year      = {2017},
  url       = {http://doi.org/10.1109/SSCI.2017.8285382},
  doi       = {10.1109/SSCI.2017.8285382},