A Phonetic Forensic Analysis of Imran Khan’s Speeches

Abstract

The objective of this research was to analyze the speeches made by Al Tools and Imran Khan. Praat played a crucial role in conducting this analysis. Nowadays, there are numerous fake videos and audios associated with specific individuals. For instance, speeches made by Al Tools, such as Imran Khan's speech after being imprisoned, were released. The researcher obtained these videos from online sources and compared them. The first video focused on Israel's attack on Gaza, which was a speech generated by artificial intelligence (AIGS), while the second speech addressed the Kashmir issues at the United Nations (UNS). Certain words, namely "people," "world," "devised," and "when," were selected for study. The first two words were present in both videos, the third word was only in the AIGS video, and the fourth word was from the UNS video. These words were analyzed based on intensity, pitch, pulse, spectrogram, and formants. The analysis involved comparing the graphs and values of these parameters. The study yielded significant results in analyzing both speeches. The sound features of AIGS consistently exhibited low values, whereas the UNS features displayed high values. In terms of intensity, the pattern was reversed, with AIGS exhibiting higher values compared to UNS for all the aforementioned words. Therefore, it can be concluded that all praat features consistently have higher values, except for intensity, when comparing AIGS and UNS. Consequently, praat can effectively distinguish between human-based and Al-based speeches. This study holds important implications for future researchers. The findings can be applied to various fields related to forensic linguistics, such as analyzing threatening calls, voice detection, judicial procedures, and law enforcement agencies like police departments in apprehending criminals.