Overview of the proposed model

  

Figure: Overview of signal-flow in proposed model.

Fig. 5 shows an overview of signal processing of the proposed model. First, the noisy vowel /a/ f(t) shown in Fig. 5 A (the SNR of f(t)is 10 dB) is decomposed into S_k(t) and $\phi_k(t)$ as shown in Figs. 5 B and C, respectively. Next, F₀(t) is estimated as shown in Fig. 5 D. The concurrent time-frequency region of the desired signal f₁(t) is determined using constraints (i), (iii), and (iv-b) as shown in Figs. 5 E, F, and G. Finally, the instantaneous amplitudes and the instantaneous phases of the two signals are determined from S_k(t) and $\phi_k(t)$ using constraints (ii) and (iv-a). The determined A_k(t) and $\theta_{1k}(t)$ are shown in Figs. 5 I and J, respectively. The segregated signal $\hat{f}_1(t)$ is shown in Fig. 5 K. In this figure, the segregated B_k(t), $\theta_{2k}(t)$, and $\hat{f}_2(t)$ are omitted.