In this paper, we proposed a computational model of CMR. This model consists of two models, our auditory segregation model (model A) and the power spectrum model of masking (model B), and a selection process that selects one of their results. The mechanisms for extracting a sinusoidal signal from a masked signal work as follows: model A uses the outputs of multiple auditory filters and model B uses the output of a single auditory filter. Simulations of the two models were carried out using two types of noise masker, the same as Hall et al.'s demonstration conditions, bandpassed random noise and AM bandpassed random noise. In model A, the signal threshold decreased depending on the type of masker and the masker bandwidth. In the case of bandpassed random noise, the signal threshold did not vary as the masker bandwidth increased. In contrast, for AM bandpassed noise, the signal threshold decreased as the masker bandwidth increased. In model B, the signal threshold increased as the masker bandwidth increased up to 1 ERB and then remained constant for both noise maskers. The selection process then selected the highest SNR from the sinusoidal signals extracted from the results of the two models. As a result, the characteristics of the proposed model show that the phenomenon of CMR is similar to Hall et al.'s results. The maximum amount of CMR in the proposed model was about 8 dB.
Hence, the proposed model can be interpreted as a computational model of CMR. It was also shown that regularity (iv) is one clue to CMR.