A simple robust fundamental frequency estimator on the time-frequency domain is proposed. Combined with the appropriately designed low-pass filter, the instantaneous frequency estimator based on the Teager-Kaiser energy function can detect the fundamental frequency of speech signal. The Teager-Kaiser function can be obtained through real computation and show the change of frequency as time goes. And when a speech block with N samples is processed with a lowpass fille. with length of L, it requires $O(N{cdot}(L+5))operations,$ compared to $O(N{cdot}2log_2N+L))operations$ in the recently introduced wavelet and conventional instantaneous frequency method. The computer simulation confirms the usefulness of the proposed fundamental frequency estimation method.