Improving beampatterns of two-dimensional random arrays using convex optimization.

Area of Science:

• Signal Processing
• Array Signal Processing
• Optimization

Background:

• Ubiquitous sensors are increasingly combined in arrays for source localization.
• Classical conventional beamforming with random sensor arrays results in poor beampatterns.

Purpose of the Study:

• To improve beampatterns of sensor arrays by pre-computing sensor weights.
• To develop a frequency-domain formulation for optimizing sensor weights.

Main Methods:

• Formulating the problem in the frequency domain with a desired look direction, a frequency-independent transition region, and minimized power in a rejection region.
• Utilizing convex optimization to obtain frequency-dependent sensor weights.
• Demonstrating the approach for real 2D sensor arrays.

Main Results:

• Pre-computed, data-independent sensor weights significantly enhance array beampatterns.
• The proposed method achieves improved beampatterns with computational complexity comparable to conventional beamforming.
• The convex optimization approach successfully yields frequency-dependent sensor weights.

Conclusions:

• Pre-computation of sensor weights is an effective strategy for improving beamforming in sensor arrays.
• The developed frequency-domain formulation and convex optimization provide a robust method for designing sensor array weights.
• This technique offers a computationally efficient way to achieve superior source localization performance.