Abstract
In this letter, we proposed a new algorithm to solve the problem concerning the direction-of-arrivals (DOAs) estimation of coherently distributed sources based on the block sparse signal matrix uncertainty model of compressed sensing. Considering the measurement matrix corrupted by the unknown noise, the central DOA and the angular spread estimation may degrade considerably. But our method is robust with the measurement matrix uncertainty. Furthermore, the proposed method has better performance in low signal-to-noise (SNR). The effectiveness of our method is confirmed by simulation results.
1. Introduction
In modern wireless communication and radar systems, accurately locating signal sources is a primary objective. While traditional models treat radio sources as single "points," real-world environments—particularly urban or suburban areas—introduce local scattering. Signals bounce off buildings, trees, and other obstacles, arriving at the receiver from a range of angles. These are known as coherently distributed (CD) sources. To effectively locate such sources, it is necessary to estimate both the central Direction-of-Arrival (DOA) and the "angular spread."
Compressed Sensing (CS) has emerged as a powerful tool for DOA estimation, allowing for high-resolution results using fewer hardware receivers than traditional methods. However, CS algorithms typically assume that the "measurement matrix"—the mathematical representation of the sensor array—is known perfectly. In practice, hardware imperfections and environmental noise introduce "measurement matrix uncertainty," which can cause standard algorithms like LASSO to become unstable and produce inaccurate results.
2. System Model and Problem Formulation
The research considers a Uniform Linear Array (ULA) consisting of $N$ sensors. To reduce cost and power consumption, the $N$ analog signals are compressed into $M$ digital signals ($M < N$) through a measurement matrix.
The core challenge addressed is that the actual measurement matrix $R$ is often corrupted by unknown noise $\Xi$, such that the observed data follows a model where the matrix is $R + \Xi$. Without accounting for this uncertainty, the estimation of the source's spatial distribution becomes highly prone to error.
3. The MU-selector Algorithm
The authors propose the Matrix Uncertainty Selector (MU-selector) algorithm to address these challenges. The method leverages several key signal processing concepts:
- Block-Sparse Modeling: Distributed sources naturally occupy a continuous cluster of angles. In the spatial domain, this translates to "block-sparsity," where non-zero signal entries appear in contiguous blocks. The MU-selector uses this structural property to improve the accuracy of the reconstruction.
- Uncertainty Integration: Unlike standard models, the MU-selector treats the measurement matrix as a variable. By incorporating the known bound of the matrix noise ($\delta$) directly into the optimization objective, the algorithm becomes robust against hardware errors.
- Dimensionality Reduction: To maintain computational efficiency, the algorithm applies Singular Value Decomposition (SVD) to the received signal matrix, focusing the optimization on the most significant signal components.
- Convex Optimization: The final DOA and angular spread are determined by solving a convex optimization problem that balances residual error, signal sparsity, and hardware uncertainty.
4. Simulation Results and Discussion
The algorithm was validated using synthetic simulations of a ULA with 20 sensors compressed down to just 5 receivers. The measurement matrix was intentionally corrupted with Gaussian noise to test robustness across various Signal-to-Noise Ratios (SNR).
Key Findings:
- Shape Recovery: Traditional methods often produce narrow, "sharp" spikes that fail to represent the physical width of a distributed source. The MU-selector successfully recovers the "block" shape of the signal, allowing for precise calculation of the angular spread.
- Robustness at Low SNR: The proposed method demonstrates significant performance gains in noisy environments. At a low SNR of 5 dB, the Root Mean Square Error (RMSE) for the central DOA was approximately half that of existing state-of-the-art methods (such as Gan’s method).
- Consistency: Through 200 Monte-Carlo simulations, the MU-selector proved to be consistently reliable across different noise patterns, confirming its practical utility.
5. Conclusion
The MU-selector algorithm provides a robust solution for the DOA estimation of coherently distributed sources under conditions of measurement matrix uncertainty. By combining block-sparse constraints with an uncertainty-aware optimization framework, the method achieves superior accuracy and robustness, particularly in low-SNR scenarios. This approach paves the way for more reliable signal localization in complex, real-world environments using cost-effective, compressed hardware.
Summary updated