Multivariate acoustic detection of small explosions using Fisher's combined probability test

Area of Science:

• Seismology and Acoustics
• Signal Processing
• Statistical Hypothesis Testing

Background:

• Identifying weak explosion signals in complex acoustic environments is challenging.
• Existing methods often struggle with highly variable background noise.
• A robust detection and discrimination methodology is needed.

Purpose of the Study:

• To develop a methodology for combined acoustic detection and discrimination of explosions.
• To identify weak explosion signals obscured by complex background noise.
• To apply a novel framework to real-world acoustic data.

Main Methods:

• Utilizing physical models for simple explosions formulated as statistical hypothesis tests.
• Employing three discriminants for signal analysis.
• Applying Fisher's Combined Probability Test to integrate p-values from multivariate discriminants.

Main Results:

• Successfully developed a methodology for acoustic explosion detection and discrimination.
• The approach effectively identifies weak explosion signals within complex noise.
• Demonstrated the framework's applicability using acoustic data from a controlled explosion.

Conclusions:

• The proposed methodology offers a robust solution for detecting and discriminating explosions acoustically.
• The approach is effective even with complex and variable background noise.
• This framework advances the capability to identify subtle explosion events in challenging acoustic settings.