Spectral technology can extract useful characteristic information from a large number of raw signals, which can be directly utilized for analyzing and identitying the material components of the observed samples. It has high application value in fields such as biomedicine, food safety and military reconnaissance. Due to the varying objectives and effects of the pretreatment, there are currently multiple spectral pre-processing methods available. We propose a spectrum signal pre-processing algorithm based on multi-scale wavelet transform, and the performance of the proposed algorithm and the designed softwere are evaluated through tests using both simulated and experimental spectra. The signal-to-noise ratio (SNR) of the simulated signal is 0.5 dB. After processing with the algorithm proposed in this paper, the SNR can reach to 8.978 dB. In the simulation, five different types of baselines are introduced, including linear, Gaussian, polynomial, exponential, and sigmoidal function types. Baseline estimation is performed using the algorithm proposed in this paper. The root mean square errors (RMSE) of the estimated values are 0.3759, 0.2883, 0.6631, 0.3489, 0.4520, respectively. The spectrum of Polytetrafluoroethylene was measured using a confocal micro-Raman spectrometer and preprocessed with the algorithm proposed in this paper. The results demonstrate that the algorithm is capable of fast and accurate processing of the spectra. The algorithm can be used to reduce noise and correct baseline. This study put on a set of new ideas on spectrum signal processing.