High-resolution broadband beamforming and detection methods with real data

Shefeng Yan, Yuanliang MaNippon 2004 · DOI

Abstract

Modern passive sonar systems face the challenge of detecting increasingly quiet targets against a background of significant ambient noise and clutter. This paper explores a high-resolution approach that combines Minimum Variance Distortionless Response (MVDR) beamforming with Subband Peak Energy Detection (SPED). By leveraging the spatial coherence of targets across frequency subbands, the proposed method achieves superior resolution and detection performance compared to conventional energy detection techniques.

Introduction

In passive sonar applications, the primary goal is to distinguish acoustic signatures of interest from background noise. Traditional Conventional Energy Detection (CED) methods often result in broad, smeared targets on sonar displays, making it difficult to resolve closely spaced sources or detect weak signals in the presence of strong interference. To overcome these limitations, high-resolution beamforming and advanced post-processing techniques are required.

Methodology

The proposed system utilizes a multi-stage processing pipeline designed to sharpen spatial resolution and enhance signal-to-noise ratio (SNR).

MVDR Beamforming

The first stage employs Minimum Variance Distortionless Response (MVDR) beamforming. Unlike standard delay-and-sum beamformers, MVDR adaptively minimizes the total output power while maintaining a constant gain in the direction of interest. This effectively places nulls in the direction of interferers, providing a higher-resolution initial spatial estimate.

Subband Peak Energy Detection (SPED)

The core innovation lies in the Subband Peak Energy Detection (SPED) technique. The broadband signal is first decomposed into multiple frequency subbands using a Short-Time Fourier Transform (STFT). While conventional methods average the energy across these subbands, SPED identifies local peaks in the azimuth spectrum for each individual frequency bin.

Architecture of the Subband Peak Energy Detection (SPED) system.

As shown in the architecture above, the "peak pick" stage exploits the fact that true targets maintain spatial coherence across different frequencies, whereas random noise does not. When these peaks are aggregated, the target signals add constructively, while noise remains suppressed, leading to much thinner and brighter tracks on the final output.

Experimental Setup

The performance of the MVDR-SPED pipeline was evaluated using both simulated and real-world data:

  • Simulation: A 24-element circular hydrophone array was modeled with two targets located at 106° and 126° azimuth, operating at a low SNR of -18 dB.
  • Real-World Data: Experimental data was collected from a 24-element circular array deployed in shallow water. The system processed 23 frequency subbands between 704 Hz and 2,112 Hz.

Results and Discussion

The simulation results clearly demonstrate the resolution advantage of the SPED approach. In the azimuth energy spectrum, the targets at 106° and 126° appear as two distinct, sharp peaks.

Azimuth energy spectrum using SPED for two simulated targets at 106° and 126°.

In contrast, conventional methods (CED) produce broad, overlapping humps that make target separation difficult. The real-world experimental data further confirmed these findings, with the SPED-processed Bearing-Time Records (BTR) showing significantly clearer and more trackable lines for noise sources. This "thin and bright" characteristic is vital for sonar operators tasked with identifying weak targets in cluttered environments.

Conclusion

The combination of MVDR beamforming and SPED offers a practical and effective solution for high-resolution broadband sonar detection. By focusing on the spatial coherence of signals across subbands through peak picking, the system significantly improves target resolution and detection sensitivity. These methods are computationally feasible for real-time implementation in modern multibeam sonar systems.

Summary updated

BibTeX
@inproceedings{yan2004high,
  author = {Shefeng Yan and Yuanliang Ma},
  title = {{High-resolution broadband beamforming and detection methods with real data}},
  booktitle = {Nippon},
  year = {2004},
  doi = {10.1250/ast.25.73},
}
Figure