Abstract: Blind source separation of convolutive mixtures based on double frequency-point permutation is proposed in this paper to address the impact of permutation indetermination on the blind source separation of convolutive mixtures. Firstly, a short-time Fourier transform is performed on the convolved signals, an instantaneous mixed model is established at each frequency point in the frequency domain for independent component analysis, and the separation results at each frequency point are replaced by the first permutation based on similar frequency points with influence factor (IF-Murata permutation). Then, a second frequency-point optimization is performed for the permutated frequency points according to the signal energy magnitude. Finally, by comparing the frequency-point correlations in the permutations, it can be seen that the double frequency-point permutation method can effectively improve the accuracy of frequency-point permutation, screen out the optimal permutation results, and further improve the signal separation performance. Simulation results show that compared with the results obtained by the IF-Murata permutation algorithm, the source-to-interference ratio, source-to-distortion ratio and source-to-artifact ratio obtained by the double frequency-point permutation method are all improved, and the similarity coefficient is improved by about 0.1 on average, which proves the effectiveness of the algorithm.