In this study, a peak detection algorithm based on an even-odd function to decompose fiber Bragg grating (FBG) spectra is proposed and experimentally demonstrated. An FBG spectrum is a function of wavelength λ and light intensity P(λ), which is moved along the λ axis point-by-point and intersected with the P(λ) axis. The maximum value of the odd function used for the FBG spectral decomposition under each moving point is defined as a characteristic value. The λ value corresponding to the minimum characteristic value is used as the central wavelength of the FBG spectrum. A theoretical model of the FBG spectrum is established using an asymmetric Gaussian model and additive white Gaussian noise superposition. Linearly increasing axial strain is imposed on an FBG sensor in turn. The central wavelengths of the theoretical model and experimental FBG spectra are calculated using the proposed algorithm. The results show that using proposed algorithm, the central wavelength of the FBG spectrum can be obtained more quickly and accurately than conventional algorithms.