Abstract: In signal processing applications such as radar and sonar, known signal detection in impulsive noise environments poses a critical challenge. Traditional methods like the matched filter often suffer from severe performance degradation when confronted with large-amplitude outliers in impulsive noise. To address this issue, a robust detection method named the Gaussian Kernel Correlator (GKC) is proposed, inspired by the mechanism of the Gaussian kernel function to effectively suppress large-valued outliers. The study employs a Contaminated Gaussian Model (CGM) to characterize impulsive noise, systematically derives closed-form expressions for the mean and variance of the GKC, and further provides approximate analytical expressions for the false alarm and detection probabilities, thereby establishing a theoretical basis for performance evaluation and threshold setting. Simulation results and theoretical analysis demonstrate that, under impulsive noise conditions, the GKC significantly outperforms the conventional matched filter and rank-based correlation detectors, and approaches the theoretically optimal locally optimal detector. Meanwhile, in additive white Gaussian noise environments, the GKC maintains detection performance comparable with that of the matched filter. This study confirms that the GKC is a novel detection method offering robustness, high performance, and practicality in complex noise environments.