An eigenvector-based test for local stationarity applied to array processing

Area of Science:

• Signal Processing
• Underwater Acoustics
• Array Processing

Background:

• Estimating data covariance matrices in sonar array processing is difficult due to limited data stationarity with moving targets.
• Accurate covariance matrix estimation is crucial for target detection and source localization in dynamic underwater environments.

Purpose of the Study:

• To develop a data-driven method for segmenting sonar data into locally stationary intervals.
• To provide improved bounds for snapshot computation in time-varying covariance matrix estimation.
• To enhance the detection of weak targets in complex acoustic scenarios.

Main Methods:

• An eigenvector-based approach is utilized for segmenting the data into locally stationary intervals.
• The method determines data-driven upper bounds for the number of snapshots required for covariance matrix computation.
• Simulated data from a horizontal array is used to validate the technique.

Main Results:

• The eigenvector method effectively segments data, identifying intervals of local stationarity.
• This segmentation provides higher, data-driven bounds for snapshot usage in covariance matrix calculations.
• The application demonstrated successful detection of a quiet source despite a loud interferer.

Conclusions:

• The proposed eigenvector-based segmentation method offers a robust solution for covariance matrix estimation in sonar array processing with moving targets.
• This approach improves the reliability of target detection by maximizing the utility of available data snapshots within stationary intervals.
• The technique shows promise for real-world applications in underwater acoustics, particularly for detecting low-power sources in noisy environments.