Pattern recognition of the polygraph using fuzzy classification

Abstract

Polygraph tests are a widely used method to distinguish between truth and deception. Polygraph charts are usually analyzed by human interpreters. However, computer algorithms are now being developed to score the tests or verify the results. These methods are based on statistical classification techniques. In this study a number of time, frequency and correlation domain features were selected and used. The fuzzy K-nearest neighbor algorithm was used to classify the polygraph charts; a correct classification of ninety-one percent was obtained for a set of one hundred case files supplied by the NSA.<>