Abstract
For spatial-frequency diversity radar, diversity channels may receive partially correlated target returns and channel signal-to-noise ratios (SNRs) may be different. For six typical scenarios of diversity radar, we design detection algorithms and analyze their detection performances in theory and via numerical results. It is shown that the detectors considering target correlation and channel SNR distribution can improve the detection performance of diversity radar. Whether certain channel SNRs can achieve a higher optimal detection probability than another depends on total channel SNR and false alarm rate.
Introduction
In Multiple-Input Multiple-Output (MIMO) radar systems, diversity is a critical mechanism for overcoming the "scintillation" effect, where a target’s radar cross-section (RCS) fluctuates, potentially rendering it invisible to single-site radars. By utilizing spatial diversity (distributed radar sites) and frequency diversity (multiple carrier frequencies), engineers can stabilize detection performance.
However, traditional models often simplify target returns as either perfectly correlated or completely independent. In real-world Spatial-Frequency Joint Diversity Radars (SFDR), signals are typically partially correlated. Furthermore, because radar sites are at varying distances from the target, the signal-to-noise ratio (SNR) is rarely uniform across channels. This paper addresses these complexities by developing detection algorithms tailored for non-uniform, partially correlated environments.
System Model and Scenarios
The study considers a distributed radar system where multiple sites observe a target from different angles using multiple frequencies. This setup creates a complex signal environment where the degree of correlation depends on the target's physical dimensions and the radar's geometry.
To evaluate detection performance comprehensively, the authors categorize radar environments into six typical scenarios based on the distribution of channel SNRs and the correlation of target returns.
The General Gaussian Signal Detector (GGSD)
The core contribution of this research is the derivation of the General Gaussian Signal Detector (GGSD). Unlike standard detectors, the GGSD utilizes a Target Fluctuation Covariance Matrix (TFCM) that incorporates:
- The "jinc" function: A mathematical tool used to model how target returns decorrelate based on target size and the "Equivalent Frequency Interval."
- SNR Distribution: A vector representing varying signal strengths across different diversity channels.
The GGSD is compared against simpler benchmarks, such as the Incoherent Accumulating Detector (IAD) for independent signals and the Coherent Accumulating Detector (CAD) for perfectly correlated signals.
Key Insights and Results
- Optimal Correlation: The "sweet spot" for signal correlation is dynamic. At high Total SNRs (TSNR), independent signals (low correlation) are easier to detect. Conversely, at low TSNRs, highly correlated signals provide better performance by allowing for more effective coherent-like integration.
- The Advantage of Unbalance: Counter-intuitively, non-uniform SNR distribution can be beneficial. If the total power is low and the false alarm rate is strict, concentrating power into a single "loud" channel can yield a higher detection probability than spreading it thinly across many channels.
- GGSD Robustness: While the GGSD is sensitive to target size estimates, simulations show it remains robust even with slight mismatches, making it viable for practical deployment where exact target dimensions are unknown.
Conclusion
By accounting for partial correlation and unequal SNR distributions, the GGSD provides a significant performance boost over traditional "one-size-fits-all" radar detectors. This research highlights that the optimal radar configuration is not static but depends heavily on the operating environment, total signal power, and required false alarm thresholds.
Summary updated