Machine learning methods for credibility assessment of interviewees based on posturographic data

Sashi K Saripalle; Spandana Vemulapalli; Gregory W King; Judee K Burgoon; Reza Derakhshani

doi:10.1109/EMBC.2015.7319932

Machine learning methods for credibility assessment of interviewees based on posturographic data

Annu Int Conf IEEE Eng Med Biol Soc. 2015:2015:6708-11. doi: 10.1109/EMBC.2015.7319932.

Authors

Sashi K Saripalle, Spandana Vemulapalli, Gregory W King, Judee K Burgoon, Reza Derakhshani

PMID: 26737832
DOI: 10.1109/EMBC.2015.7319932

Abstract

This paper discusses the advantages of using posturographic signals from force plates for non-invasive credibility assessment. The contributions of our work are two fold: first, the proposed method is highly efficient and non invasive. Second, feasibility for creating an autonomous credibility assessment system using machine-learning algorithms is studied. This study employs an interview paradigm that includes subjects responding with truthful and deceptive intent while their center of pressure (COP) signal is being recorded. Classification models utilizing sets of COP features for deceptive responses are derived and best accuracy of 93.5% for test interval is reported.

Publication types

Controlled Clinical Trial
Research Support, Non-U.S. Gov't

MeSH terms

Adult
Algorithms
Discriminant Analysis
Female
Humans
Machine Learning*
Male
Posture / physiology*
Surveys and Questionnaires
Young Adult