2-D Unitary ESPRIT-Like Direction-of-Arrival (DOA) Estimation for Coherent Signals with a Uniform Rectangular Array

Shiwei Ren, Xiaochuan Ma, Shefeng Yan, Chengpeng Hao2013 · DOI

TL;DR

This paper introduces a novel 2-D Unitary ESPRIT-Like (2-D UESPRIT-like) algorithm for estimating the direction of arrival (DOA) of coherent signals using a uniform rectangular array. The key innovation is a real-valued processing approach that significantly reduces computational complexity compared to existing methods, while maintaining high estimation accuracy and robustness.

What Problem Does This Paper Address?

Estimating the direction of arrival (DOA) of multiple signals, especially when they are coherent (i.e., their signals are highly correlated), is a fundamental challenge in various fields like radar, wireless communications, and sonar. Coherent signals can lead to a rank deficiency in the covariance matrix, making traditional DOA estimation techniques perform poorly. Existing methods like spatial smoothing can decorrelate signals but require computationally expensive peak searching. Other advanced techniques, such as ESPRIT-like algorithms, while offering better accuracy, still suffer from heavy computational burdens due to complex-valued matrix operations like eigenvalue decomposition (EVD) and singular value decomposition (SVD). This paper aims to bridge this gap by developing a computationally efficient and accurate method for 2-D DOA estimation of coherent signals.

Methods

The proposed 2-D UESPRIT-like algorithm builds upon existing ESPRIT-like frameworks but introduces a crucial real-valued processing step. The core methodology involves:

  • Decorrelation of Coherent Signals: A block Hankel matrix is constructed from the covariance matrix. This matrix, when processed using a forward-backward averaging-like method, effectively decorrelates the coherent signals.
  • Unitary Transformation for Real-Valued Processing: The algorithm employs a 2-D unitary transformation (specifically, a Kronecker product of 1-D unitary matrices) to convert complex-valued computations into real-valued ones. This is a significant departure from previous methods that relied on complex EVD and SVD.
  • 1-D ESPRIT for DOA Extraction: By transforming the problem into a real-valued domain, the 2-D DOA estimation is simplified to performing 1-D unitary ESPRIT on derived matrices. This avoids the need for complex 2-D matrix operations.
  • Computational Simplification: The use of real-valued matrices and 1-D operations drastically reduces the computational complexity, making the algorithm more practical for real-time applications.

Datasets

Not discussed in the paper. The paper relies on simulated data for its evaluation.

Key Insights and Ideas

  • The Power of Real-Valued Processing: The most significant insight is that by carefully applying unitary transformations, complex-valued DOA estimation problems can be effectively solved using real-valued arithmetic. This not only simplifies the mathematics but also leads to substantial computational savings.
  • Decorrelation is Key for Coherent Signals: The paper emphasizes the importance of robust decorrelation techniques, like the forward-backward averaging applied to the block Hankel matrix, for handling coherent signals.
  • Bridging 1-D and 2-D ESPRIT: The method elegantly extends the benefits of 1-D Unitary ESPRIT to the 2-D scenario, leveraging the structure of the uniform rectangular array.
  • Computational Efficiency without Sacrificing Accuracy: The proposed algorithm demonstrates that high estimation accuracy can be achieved even with significantly reduced computational load, making it a more attractive option for practical deployment.

Key Results

  • Significant Computational Savings: The paper quantifies the computational complexity and shows that the 2-D UESPRIT-like method offers substantial savings in real multiplications compared to the 2-D ESPRIT-like method, especially as the array dimensions (P and Q) increase. For example, the savings are estimated to be around 15(PQ)³ in certain conditions.
  • High Estimation Accuracy: Simulation results demonstrate that the proposed method achieves high DOA identification probabilities and root mean square errors (RMSE) that are close to the Cramer-Rao bound (CRB).
  • Superior Performance for Coherent Signals: The 2-D UESPRIT-like method outperforms traditional 2-D spatial smoothing and is comparable to or slightly better than the 2-D ESPRIT-like method, particularly in scenarios with highly correlated signals.
  • Robustness to SNR: The algorithm maintains good performance across a wide range of Signal-to-Noise Ratios (SNRs).

Limitations and Future Work

  • Performance Degradation with Uncorrelated/Partly Correlated Signals: The paper notes that while the algorithm excels with highly correlated signals, its performance might degrade to some extent when signals are uncorrelated or only partly correlated.
  • Assumption of No Noise (for derivation): The derivation of the block Hankel matrix expression (Equation 11) is performed under the assumption of no noise. However, the paper states that simulations will address the case with spatially white noise.
  • Future Work: The authors suggest future research directions including:
  • Extending the real-valued processing of UCA ESPRIT to leverage the special structure of unitary matrices.
  • Investigating the application of the method to uniform circular arrays (UCAs) in the presence of mutual coupling.
  • Addressing the elevation-dependence of mutual coupling matrices.

Summary updated

BibTeX
@misc{ren2013d,
  author = {Shiwei Ren and Xiaochuan Ma and Shefeng Yan and Chengpeng Hao},
  title = {{2-D Unitary ESPRIT-Like Direction-of-Arrival (DOA) Estimation for Coherent Signals with a Uniform Rectangular Array}},
  year = {2013},
  doi = {10.3390/s130404272},
}

Loading related content...

Figure