Pilot Study: Polygraph Decision Support System Using Event Resolution Imaging for the Relevant/Irrelevant Format.

摘要

A mathematical and statistical signal processing strategy, termed Event Resolution Imaging (ERI), was applied to the problem of developing a new polygraph decision support system for the Relevant/Irrelevant question format. The Event Resolution Imaging (ERI) methodology was employed to analyze and separate raw polygraph signals into truthful and deceptive categories. By extracting the traditional point features of respiratory waveform length, cardiovascular baseline amplitude and EDR amplitude, along with the signal features of distinct frequencies, latencies, amplitudes and phase correlations, a relatively high accuracy has been achieved. Scoring protocols were designed using data from 577 polygraph examinations for which examinee veracity was known. The optimal scoring protocol was cross-validated using 144 polygraph examinations for which examinee veracity was not known. The veracity of examinee responses to the question 'Have you smoked marijuana in the last 30 days.' was correctly identified for 109 (i.e., 76%) of the 144 cases. This accuracy was 13% higher than that achieved by a commercial scoring algorithm and 6% higher than the estimated accuracies achieved by human experts.