A precise self-adaptive multi-peak detection algorithm is proposed in this paper, which is used to process the spectral signal of the distributed fiber Bragg grating (FBG) sensor systems. The wavelet threshold de-noising method is used to process the original spectral signal, and the Hilbert transform and the Gabor filtering method are introduced for self-adaptive segmentation of the peak region of the multi-peak spectrum. Taking the asymmetric characteristics of spectral peak into consideration, the shift of each peak is determined by the left and right half-peak areas. In addition, the exponentially modified Gaussian function is designed to modify the peak position. The experiments and simulation results show that the proposed algorithm performs better than the traditional algorithms in detection precision and stability for the multi-peak spectral signal. This implies that the proposed algorithm provides a precise demodulation algorithm for the distributed FBG sensor network, and the multi-peak position can be achieved adaptively.