[Purposes] This study addresses the challenges of anomaly detection and environmental noise interference in bridge monitoring signals by proposing an enhanced boxplot algorithm based on the least squares method for anomaly detection and replacement, alongside a denoising technique that integrates Fast Ensemble Empirical Mode Decomposition (FEEMD) with wavelet thresholding.[Methods] Initially, the improved boxplot algorithm is utilized to identify and replace anomalies within the bridge monitoring signals. Following this anomaly processing, FEEMD is applied to decompose the signal. The variance contribution rate of the decomposition components is calculated, leading to the removal of components with lower variance contribution rates. Subsequently, the remaining signal undergoes wavelet thresholding, and the denoised signal is reconstructed.[Findings] The analysis of simulated signals and real measured signals shows that the improved box plot algorithm based on the least squares method achieves a recall rate of 100% in outlier recognition. Additionally, the denoising method combining FEEMD with wavelet thresholding effectively filters out interference noise signals. The analysis of the denoising results for simulated signals with different noise levels indicates that the root mean square error (RMSE) after denoising is the smallest compared to wavelet thresholding, FEEMD, and EMD combined with wavelet thresholding, with values of 0.4676, 0.2619, 0.16195, and 0.1034, respectively. Furthermore, the signal-to-noise ratio (SNR) is improved by 8 dB to 12 dB, with values of 15.81 dB, 21.63 dB, 25.64 dB, and 28.50 dB, respectively. The results of processing real measured signals also show that the smoothness index values and the residual noise levels after denoising are superior to those achieved by wavelet thresholding, FEEMD, and EMD combined with wavelet thresholding denoising methods. [Conclusions] The research results show that the proposed improved box plot algorithm based on the least squares method and the denoising method combining FEEMD with wavelet thresholding can effectively address the issues of outlier recognition and noise in bridge monitoring signals. The research findings can serve as a reference for the preprocessing of bridge signals.