Time difference of arrival estimation for signals is an important problem in communications. The performance of conventional time delay estimation methods degrade severely in the presence of interfering signals and noise. By exploiting cyclostationarity property of cyclostationary signals, a class of signal-selective time difference of arrival (TDOA) estimation algorithm is proposed. In order to enhance the performance of signal-selective methods, a modified cyclostationarity-based TDOA estimation algorithm is proposed for cyclostationary signals. The proposed method is derived as a modification of the generalized spectral coherence alignment method (G-SPECCOA). Simulation results show that the proposed algorithm outperforms the G-SPECCOA method in the presence of Gaussian noise and interference.