Results and discussion

  

Figure: Relationship between the bandwidth related to the number of adjacent auditory filters and the SNR for the extracted signal $\hat{f}_{1,A}(t)$.

Simulations were carried out according the conditions mentioned above. The results are shown in Fig. , where the vertical and horizontal axes show the improved SNR of the extracted sinusoidal signal $\hat{f}_{1,A}(t)$ and the bandwidth related to L, respectively. Moreover, the real line and the error bar show the mean and standard deviation of the SNR of the signal $\hat{f}_{1,A}(t)$ extracted from 300 mixed signals, respectively. It was found that for the mixed signal f_M(t), a sinusoidal signal $\hat{f}_{1,A}(t)$ became detectable as the number of the adjacent auditory filters L increased, but for the mixed signal f_R(t), $\hat{f}_{1,A}(t)$ was not detectable as L increased. Therefore, the results show that a sinusoidal signal is more detectable when the components of the masker have the same amplitude modulation pattern in different frequency regions or when the fluctuations in the masker envelopes are coherent. Hence, model A simulates the phenomenon of reduction from masking using the outputs of multiple auditory filters.