We used the root-mean-squared (RMS) error between the reference F0(t) pattern and the estimated F0(t) to evaluate the estimation performance of the fundamental frequency F0(t). The reference pattern was extracted from clean speech using TEMPO [5]. The RMS errors of F0(t) for simulations 1, 2, and 3 were 3.8, 1.8, and 1.1 Hz, respectively. The results show that the proposed method could estimate F0(t) with high accuracy.