Abstract: To meet the requirements of high-accuracy and real-time frequency estimation in high-speed digital demodulators, this paper analyzes the characteristics of subspace decomposition algorithms and proposes a novel high-accuracy frequency estimation algorithm that the root multiple signal classification ( root-MUSIC) algorithm and the estimation of signal parameters via rotational invariance techniques (ESPRIT) combined with fast Fourier transform (FFT) algorithm, respectively. Firstly, the root-MUSIC or the ESPRIT algorithm is applied to pre-estimate the signal frequency; then, the narrowband spectrum interested is determined; finally, the Hanning window with FFT algorithm is used to complete the accurate frequency search by the method of frequency refinement. In addition, in order to satisfy the conditions of different signal to noise ratios (SNRs) for high-accuracy frequency estimation, a new algorithm based on the optional construction is proposed. Simulation results show that not only is the estimation accuracy greatly improved compared to pre-existing algorithms because of notable anti-noise performance but also the processing time of the demodulator is reduced.