A radio extraction algorithm of steady state based on the high order spectrum and time-domain analysis is presented for the radio fingerprint features extraction problem on the condition of steady-state operation. Firstly, the mathematical model of steady-state operation is constructed and the disadvantages of existing bispectrum feature extraction algorithm are analyzed. Secondly, combining the periodicity of rectangular integral bispectrum with the time-domain analysis, a modified radio extraction algorithm of steady state is proposed, whose applicability for any kind of high order spectral features extraction is also proved. Finally, the results of computation experiment with measured data verify the good efficiency and reliability of the proposed algorithm. By comparison with the traditional rectangular integral bispectrum algorithm, the proposed algorithm improves the accuracy rate from 90% to 97%. Moreover, in the case of the same accuracy rate, the proposed algorithm obtains a higher efficiency than the traditional one.