A wavelength detection method based on gated recurrent unit (GRU) network was proposed to demodulate the central wavelength of overlapping spectra in fiber Bragg grating (FBG) sensor networks. The proposed method transformed the wavelength demodulation problem of overlapping spectra in FBG into a regression problem and considered the sequence and spectrum characteristics of the spectral data. To learn the spectral data characteristics and train to achieve the corresponding wavelength detection model, the GRU network was used. Thus, the overlapping spectra could be quickly and accurately demodulated. Experimental results show that the proposed method can overcome the precise demodulation problem of the central wavelength for partially or completely overlapping spectra of FBG sensor networks. The test results with root mean square less than 1 pm account for 88.2% of the test results. The detection accuracy and stability of the proposed method provide enhanced results compared with existing demodulation methods. The proposed method provides a novel way to improve the multiplexing capability of FBG sensor networks.