An audio tampering detection framework for low SNR conditions based on modified CZT

Extracting the electric network frequency (ENF) from digital audio signals is a crucial means of forensic evidence. However, ENF signal extraction is susceptible to noise, making it challenging to establish a reliable matching relationship with the reference frequency database, especially under low signal-to-noise ratio (SNR) conditions. To solve this problem, an ENF extraction and tampering detection framework (ENF-ETD) for low SNR conditions is proposed in this article. Firstly, a modified Chirp Z-transform (MCZT) method is proposed to extract the ENF signal in digital audio. Subsequently, by comparing with the actual grid frequency, the Pearson correlation coefficient (PCC) and Euclidean distance (ED) are used to evaluate the accuracy of ENF estimation and determine whether the audio has been tampered with. Finally, the simulations and hardware-based experiments verify the proposed ENF-ETD framework's effectiveness in noise immunity and digital audio tampering detection.