<![CDATA[Abstract - This study aims to statistically analyze the parameters of acoustic phonetic features, namely: pitch, formant and intensity as indicators of psychological pressure from a person's lying or honest voice. The strength of this research is using the "true experimental" model and assessing a person's lying / honest voice in real time parallel to the polygraph test. The study was conducted on 6 subjects (3 couples) which were divided into 2 groups, namely the control group that was not treated, and the treatment group who received treatment with instructions to take money in a wallet on the table and was asked to lie not to take the money in question. . Voice samples are the answers to “No” to the relevant and comparison questions on the examined subjects on the polygraph test. The Praat software was used for feature extraction and data analysis using SPSS with the Paired Samples T Test and the Wilcoxon Sign Rank Test for different tests. Based on the results of the statistical analysis of acoustic phonetic features, a scoring system is proposed by setting a score (+1) when the results of Ho's statistical analysis are accepted (Sig value> 0.05) or in other words there is no difference between the formant, intensity and pitch values of the answer to the word "no" to the comparison question. with the answer to the word "no" in the relevant question. And vice versa set a score (-1) when the results of statistical analysis Ho is rejected (Sig value <0.05) or in other words there are differences in the value of formant, intensity and pitch of the answer to the word "no" to the comparison question with the answer to the word "no" to the relevant question. The final conclusion is that the total score (+) indicates honest and the total score (-) indicates a lie. Based on the total value of statistical analysis, the formant features are known to be very significant for detecting someone's lies, because in this study they have a 100% probability of successfully detecting lies, while the intensity and pitch features are less significant for detecting lies because they only have a 66.666% probability of successfully detecting lies 33.333%.]]>
