Abstract: The problem of time delay estimation (TDE) for cyclostationary signals in the presence of ?-stable distribution impulsive noise and interfering signals was addressed. Since the conventional algorithms based on the second-order cyclic spectral correlation function degenerated severely in impulsive noise environment, a novel robust algorithm based on the fractional lower-order spectrum was proposed. Compared with conventional spectra analysis methods, this novel method could suppress Gaussian noise, non-Gaussian impulsive noise and interfering signals more effectively. Simulation results demonstrate the effectiveness and robustness of the novel algorithm.