Acoustic environment identification by Kullback-Leibler divergence

Area of Science:

• Digital Forensics
• Acoustic Analysis
• Signal Processing

Background:

• Digital audio recordings contain multiplicative (device-related) and additive (environment-related) noise.
• Distinguishing the origin of audio recordings is crucial in forensic investigations.
• Existing methods may not sufficiently differentiate between recording environments.

Purpose of the Study:

• To develop a forensic methodology for determining the probable recording location of disputed digital audio.
• To estimate a likelihood rate to assess the plausibility of different recording environments.

Main Methods:

• Analyzing additive noise as an external feature indicative of the recording environment.
• Estimating additive noise using the Geometric Approach to Spectral Subtraction (GA-SS) filter.
• Calculating a likelihood rate via Kullback-Leibler divergence for statistical comparison of noise distributions.

Main Results:

• The methodology provides a quantitative likelihood rate to support the determination of the recording place.
• Statistical comparison of additive noise probability distribution functions aids in identifying the most plausible location.
• Reference recordings from potential locations are used for comparative analysis.

Conclusions:

• The proposed forensic methodology effectively utilizes additive noise analysis to determine the probable origin of digital audio recordings.
• This approach offers a robust tool for forensic audio investigations by quantifying environmental acoustic signatures.