In this study, a novel single-tone-based channel estimation method is proposed for orthogonal time-frequency-space (OTFS) systems. The channel parameters, including the delay taps, Doppler taps, and channel gains, are sequentially estimated in a delay-Doppler domain. Concretely, the delay taps are estimated using a threshold method that provides a constant probability of false alarms (Neyman-Pearson criterion). The integer and fractional Doppler taps and the channel gains of each separable multipath component are obtained using a single-tone parameter estimation method. The estimated parameters are then used to reconstruct the channel matrix. The simulation results demonstrate that the proposed method outperforms conventional low-complexity channel estimation methods in terms of the normalized mean squared error of the channel estimation and the bit error rate (BER) of the communications at the cost of a marginal increase in computational complexity. Further, the BER performance of the proposed method is almost identical to that of perfect channel estimation.