Fixed-point deformation instruments(FPDI), which observation precision is extremely high, yet are susceptible to noise generated by variations in the operational status and external environmental changes, resulting that seismic information is concealed in various interference signals. To eliminate noise from FPDI, a denoising model combining the improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) and the synchrosqueezing transform (SST) was proposed. According to this model, the FPDI signal containing noise was firstly decomposed into several intrinsic mode function (IMF) components through ICEEMDAN. And then, the effective IMFs and noise-containing IMFs were distinguished by considering sample entropy (SE), the variance contribution rate and correlation coefficient. Finally, SST was utilized to denoise the noise-containing IMFs, which were then reconstructed with the effective IMFs to obtain a purified signal. The simulation experiments and the analysis of different types of measured signals from the Yichang seismic station show that the ICEEMDAN-SST model can effectively distinguish the noisy components from the effective components. The signal restoration degree after denoising is relatively high, and the solid tide pattern is clear. The denoising effect is superior to traditional methods such as S-G (Savitzky-Golay) filtering, Kalman filtering, and wavelet denoising, and is suitable for suppressing different types of noise in various FPDI. The proposal of the ICEEMDAN-SST model is of great significance for the extraction of seismic information by FPDI and helps such observation instruments play a greater role in seismic analysis and prediction.