Abstract: In hydraulic engineering, sluices are critical flood discharge structures. Accurately acquiring their flow-induced vibration response signals is essential for structural health monitoring and operational safety assessment. However, field monitoring signals are vulnerable to multi-source noise interference (including low-frequency water flow, vibration of electromechanical equipment, traffic loads, etc.), which causes the dynamic characteristics of flood discharge sluice structures to be obscured by noise. Therefore, denoising of flow-induced vibration response signals is a key prerequisite for accurately obtaining the dynamic characteristics of flood discharge sluices. To address the challenge that such signals are frequently contaminated by background white noise and low-frequency flow noise, thereby affecting the accuracy of modal analysis, this study proposes a combined denoising method based on Singular Spectrum Analysis and improved Continuous Variational Mode Decomposition. First, Singular Value Decomposition is applied to calculate the contribution rates of the obtained singular values. Matrix components whose cumulative contribution rate exceeds a predefined threshold are selected to separate the effective signal matrix from the noise interference matrix, thereby filtering out background white noise. Then, the Newton-Raphson optimization algorithm is adopted to adaptively determine the quadratic penalty factor, a key parameter in Continuous Variational Mode Decomposition. Since the selection of the fitness function plays a crucial role in the optimization of Continuous Variational Mode Decomposition parameters using the Newton-Raphson-Based Optimization algorithm, this paper adopts envelope entropy as the fitness function and performs parameter optimization by calculating fitness values to evaluate the distance between individuals and the optimal solution, thereby improving the adaptive decomposition capability of the Continuous Variational Mode Decomposition algorithm. This enables precise separation of the sluice vibration signal and alleviates mode mixing. Finally, to effectively distinguish useful components from noise interference in the signal, the linear correlation degree is quantified by calculating the Pearson correlation coefficient between each Intrinsic Mode Function component and the original signal, and a threshold is set to screen the decomposed components. When the Pearson correlation coefficient between an Intrinsic Mode Function component and the original signal is lower than the preset threshold, the corresponding Intrinsic Mode Function component is eliminated, thereby removing low-frequency flow noise. Accordingly, by combining the advantages of Singular Spectrum Analysis and improved Continuous Variational Mode Decomposition, effective filtering of background white noise and low-frequency water flow noise in vibration signals of flood discharge sluices is achieved. The effectiveness of the proposed method is verified using numerical signals. The results show that, compared with denoising methods such as Singular Spectrum Analysis, Complete Ensemble Empirical Mode Decomposition with Adaptive Noise, and Singular Spectrum Analysis–Complete Ensemble Empirical Mode Decomposition with Adaptive Noise, the method proposed in this paper exhibits superior denoising performance in terms of Signal-to-Noise Ratio and Root Mean Square Error. The proposed method can increase the Signal-to-Noise Ratio of noisy signals by up to 4.61 times after denoising and reduce the Root Mean Square Error by 90.2%, demonstrating its effectiveness in vibration signal denoising. The method is further applied to the denoising of actual flow-induced vibration response signals of sluices. The results indicate that the proposed method can effectively filter out low-frequency water flow noise and background white noise. The frequency content of the denoised signal is basically consistent with the modal identification results, and it can effectively reduce pseudo-stable points and false modes in the stability diagram during operational modal parameter identification, providing strong support for the subsequent identification of structural dynamic characteristics and evaluation of operational status. In addition, with appropriate modifications, the proposed method can also be applied to the denoising of flow-induced vibration response signals and the accurate extraction of dynamic characteristics for other discharge structures such as arch dams, gravity dams, and guide walls.