A Novel Audio Forgery Detection Method Based on Short-time Power Spectral Density
Jiayi Cheng, Xiaolong Li · 2024
The popularity of portable recording equipment has made the acquisition of digital audio convenient, and the development of multimedia editing software has made audio editing and modification progressively easier. Therefore, it is essential to ensure the authenticity and integrity of digital media files. Copy-move forgery is a prevalent form of audio tampering, but existing detection methods suffer from low accuracy and poor robustness. To address this issue, we propose a detection method for audio copy-move forgery based on short-time power spectral density (STPSD). Specifically, we first separate the syllables in the voiced part of the speech signal. Then, the STPSD features are extracted from the suspect byte pair. Finally, the content matching algorithm is used for detection. The experimental results show that the proposed algorithm can resist all kinds of common attacks reliably, and can accurately locate copied and moved forged syllables, with high efficiency.