Abstract
In this paper, we propose a beam space conversion (BSC)-based approach to achieve a single near-field signal localization under a uniform circular array (UCA). By employing the centro-symmetric geometry of the UCA, we apply BSC to extract the two-dimensional (2-D) angles of the near-field signal in Vandermonde form, allowing for azimuth and elevation angle estimation via an improved estimation of signal parameters via rotational invariance techniques (ESPRIT) algorithm. By substituting these angles into the direction vector, the range parameter is obtained using a 1-D multiple signal classification (MUSIC) method. Simulations demonstrate that the proposed algorithm provides satisfactory performance with reduced computational complexity.
Introduction
Signal localization is a fundamental task in array signal processing. While far-field sources are modeled as plane waves, near-field sources—common in indoor detection, underwater acoustics, and short-range radar—produce spherical waves. This necessitates the estimation of three parameters: azimuth, elevation, and range.
Traditional methods like 3D-MUSIC offer high resolution but suffer from extreme computational costs due to exhaustive multi-dimensional grid searches. Uniform Circular Arrays (UCAs) are preferred for their 360-degree coverage, but their non-linear geometry complicates the mathematical processing compared to linear arrays. This paper introduces a method to bridge this gap by transforming circular array data into a more efficient format.
Proposed Methodology
The proposed algorithm addresses these challenges by decoupling the angular and range parameters and transforming the UCA data into a computationally friendly "Vandermonde" format.
The process involves four key stages:
- Parameter Decoupling: Exploiting the centro-symmetric geometry of the UCA to isolate angular information from the range using a correlation function.
- Beam Space Conversion (BSC): This core innovation mathematically transforms the circular array data into a Vandermonde form, effectively making the UCA behave like a virtual linear array.
- Improved ESPRIT: Calculating azimuth and elevation angles directly using closed-form matrix operations, avoiding exhaustive searches.
- 1D-MUSIC for Range: Substituting the estimated angles into the steering vector to reduce the range estimation to a simple one-dimensional search.
Experimental Results
The algorithm was evaluated using Monte-Carlo simulations with a 16-sensor UCA (1m radius) and a 300 MHz signal. Performance was measured via Root Mean Square Error (RMSE) across 500 independent trials.
Key findings include:
- Computational Efficiency: The proposed method is significantly faster than traditional approaches. In tests (SNR 20 dB, 2000 snapshots), it required only 0.73 seconds, compared to 2.12 seconds for 3D-MUSIC and 1.52 seconds for the Two-Stage MUSIC (TSMUSIC) method.
- Accuracy: As shown in the RMSE plots, the proposed algorithm's accuracy converges to that of the computationally intensive 3D-MUSIC at high Signal-to-Noise Ratios (SNR).
- Robustness: The method maintains high resolution probability even at lower SNR levels and snapshot counts compared to simpler phase-based techniques.
Conclusion
By bridging the gap between complex circular geometries and efficient linear-style algorithms through Beam Space Conversion, this method offers a high-speed solution for 3D near-field localization. While currently optimized for single-signal scenarios, its computational advantages make it a promising candidate for real-time applications in short-range sensing. Future work will focus on extending this framework to multi-signal environments and addressing sensor non-idealities.
Summary updated