Abstract: The extraction of demodulation spectrum using classical demodulation methods faces challenges such as difficulty in selecting the demodulation center frequency and bandwidth, and the interference of out-of-band noise in distinguishing the demodulation spectrum. This study proposes an approach that integrates blind deconvolution and chaotic oscillator techniques to efficiently minimize out-of-band noise and enhance the demodulation spectrum. The utilization of the Duffing oscillator is employed as a means to pre-detect low-frequency weak periodic signals inside the wideband ship radiated noise. Following this, the associated frequency is utilized as the prior parameter for the minimum noise amplitude deconvolution (MNAD) technique. The MNAD approach is employed to perform an adaptive search for the demodulation frequency band, resulting in the successful attenuation of noise signals that exist beyond the demodulation band. Ultimately, the demodulation spectrum is achieved with clarity and distinctiveness, accompanied by high signal-to-noise ratios. The results obtained from simulations and measured data prove that this method outperforms classical demodulation methods and other blind deconvolution methods in terms of gaining center frequencies and filtering bandwidths. The demodulation spectrum obtained by this method demonstrates the highest narrowband signal-to-noise ratio, as indicated by the DF value.