The spectrum of optical fiber surface plasmon resonance (SPR) shows the characteristics of wider full width at half-maximum (FWHM) and smaller peak value. The traditional peak detection algorithm cannot calculate the resonance wavelength of this kind of spectrum accurately. Therefore, efficient methods to calculate the resonance wavelength are required. This paper presents a method to process the SPR signal based on the Gaussian fitting. Combined with the trust region algorithm to determine the Gaussian fitting function, and then the resonance wavelength be calculated by line search method. Data processing of standard spectrum of glycerin with different concentration proves that this algorithm can adapt to optical fiber SPR spectrum characteristics, and calculate resonance wavelength accurately. The experiment measures the SPR signals of sucrose under different concentrations through the self-building SPR measure system. By using the proposed method, the weighted centroid method and the tracking centroid method to calculate the resonance wavelength, the result shows that the Gaussian fitting method can effectively improve resolution and have a short operation time which is conducive to engineering application.