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

Area of Science:

• Digital Signal Processing
• Forensic Science
• Electrical Engineering

Background:

• Extracting electric network frequency (ENF) from digital audio is vital for forensic evidence.
• Noise and low signal-to-noise ratio (SNR) conditions complicate reliable ENF signal extraction and audio tampering detection.

Purpose of the Study:

• To propose an ENF extraction and tampering detection framework (ENF-ETD) specifically designed for low SNR conditions.
• To enhance the accuracy and reliability of ENF extraction and audio tampering identification in challenging acoustic environments.

Main Methods:

• A modified Chirp Z-transform (MCZT) method is employed for ENF signal extraction from digital audio.
• Pearson correlation coefficient (PCC) and Euclidean distance (ED) are utilized to assess ENF estimation accuracy.
• Comparison with actual grid frequency is performed for tampering detection.

Main Results:

• The proposed ENF-ETD framework demonstrates effective ENF signal extraction under low SNR conditions.
• The framework successfully identifies audio tampering by comparing estimated and actual grid frequencies.
• Simulations and hardware experiments validate the framework's robustness against noise.

Conclusions:

• The ENF-ETD framework provides a reliable solution for ENF extraction and tampering detection in noisy digital audio.
• The MCZT method improves ENF extraction accuracy, enhancing forensic analysis capabilities.
• The study confirms the framework's effectiveness in noise immunity and digital audio tampering detection.