This paper proposed a new method of extracting the desired speech from noisy speech, by improving two aspect of our previously proposed method. One was to incorporate a method of estimating the fundamental frequency F0(t), by using the Comb filtering on the filterbank. The other was to reconsider the constraints on the separation block, which are the instantaneous amplitude, input phase, and fundamental frequency of the desired signal.
As an example of segregation using the proposed method, we demonstrated three simulations of segregating two acoustic sources. The results for extracting F0(t) showed that the proposed method can estimate the fundamental frequency with high accuracy. In particular, the results of simulations 1 and 2 examined how to solve the problem of segregating two acoustic sources. The results of simulation 3 showed that the proposed method could also extract the speech signal from noisy speech. Moreover, comparisons under three conditions showed that using the proposed method with three conditions related to the four regularities was better than using the other method under all three conditions.
 |
