TL;DR
This paper introduces a novel sparse array configuration called the Unfolded Coprime Linear Array with Three Subarrays (UCLATS) designed for Direction of Arrival (DOA) estimation of non-Gaussian signals. They also propose an efficient DFT-MUSIC algorithm that achieves comparable DOA accuracy to traditional methods but with significantly reduced computational complexity.
What Problem Does This Paper Address?
Accurate estimation of the direction of arrival (DOA) of signals is crucial in many applications like wireless communication, radar, and navigation. Traditional methods often rely on uniform linear arrays (ULAs), which require a high density of sensors to avoid angle ambiguity, limiting accuracy. Sparse arrays, like coprime arrays (CA) and nested arrays (NA), offer a solution by using fewer sensors to achieve a larger effective aperture and higher degrees of freedom (DOF). However, most existing sparse array designs are optimized for Gaussian signals. Real-world signals are often non-Gaussian, and their DOA estimation requires different techniques, typically involving higher-order statistics like fourth-order cumulants (FOC). Furthermore, existing sparse array designs often optimize subarrays independently, leading to suboptimal overall performance. This paper addresses the need for sparse array designs specifically tailored for non-Gaussian signals and proposes a systematic method to optimize the entire array configuration for maximum consecutive DOF.
Methods
The paper proposes a two-step approach for designing the UCLATS configuration:
- Global Postage-Stamp Problem (GPSP): The problem of finding the optimal placement of physical sensors to achieve the longest possible consecutive virtual sum co-array (2-SC) is modeled as a GPSP. The solution to this GPSP provides the locations for the virtual 2-SC and an initial physical sensor array.
- Coprime Coefficient Multiplication: The virtual 2-SC locations are then multiplied by appropriate coprime coefficients to generate the UCLATS configuration. The initial physical array locations are similarly scaled to obtain the final physical sensor positions.
For DOA estimation, a novel DFT-MUSIC algorithm is introduced. This method combines:
- Discrete Fourier Transform (DFT): Used for initial DOA estimates by exploiting the consecutive virtual elements of the array.
- Partial Spectrum Searching MUSIC (PSS-MUSIC): Used for fine-tuning the DOA estimates, leveraging the initial estimates to significantly reduce the search space compared to traditional Total Spectrum Searching MUSIC (TSS-MUSIC).
Datasets
Not discussed in the paper. The evaluation is performed through numerical simulations.
Key Insights and Ideas
- Systematic Sparse Array Design: The GPSP framework provides a principled way to design sparse arrays that maximize consecutive DOF, moving beyond ad-hoc subarray optimizations.
- UCLATS for Non-Gaussian Signals: The proposed UCLATS configuration is specifically designed to leverage higher-order statistics (FOC) for improved DOA estimation of non-Gaussian signals.
- DFT-MUSIC Efficiency: Combining DFT for initial estimates with PSS-MUSIC for refinement offers a significant computational advantage over TSS-MUSIC without sacrificing asymptotic performance. This is achieved by intelligently narrowing the search range for the MUSIC algorithm.
- Consecutive DOF is Key: The paper emphasizes the importance of consecutive DOF for efficient DOA estimation, particularly when using methods like DFT. The proposed design aims to maximize this property.
Key Results
- The proposed UCLATS configuration, designed using the GPSP method, demonstrates superior consecutive DOF compared to other state-of-the-art sparse arrays like ACA, SAFE-CPA, and FL-NA.
- The DFT-MUSIC algorithm achieves the same asymptotic DOA estimation performance as the traditional TSS-MUSIC algorithm.
- Crucially, the DFT-MUSIC algorithm requires less than 10% of the complex multiplications compared to TSS-MUSIC, making it significantly more computationally efficient.
- Simulation results show that the RMSE of DOA estimation improves with an increasing number of snapshots and sensors, as expected.
Limitations and Future Work
- Computational Complexity of FOC Matrix: While the DFT-MUSIC algorithm is efficient for spectrum searching, the initial computation of the FOC matrix can still be computationally intensive, especially for a large number of sensors and snapshots.
- Mutual Coupling: Although sparse arrays generally reduce mutual coupling, it's not explicitly addressed as a primary focus for the UCLATS design in this paper.
- Real-world Signal Characteristics: The paper focuses on non-Gaussian signals and their FOC properties. Future work could explore the array's performance with more complex signal models, such as correlated sources or non-stationary signals.
- Experimental Validation: The results presented are based on numerical simulations. Real-world experimental validation would further strengthen the findings.
Summary updated