A discrete-cepstrum based spectral-envelope estimation scheme with improvements

Approximating spectral envelope with regularized discrete-cepstrum coefficients was proposed by previous researchers. In this paper, we study two problems encountered in practice when adopting this approach. The first is which spectral peaks should be selected, and the second is what frequency axis scaling function should be adopted. After some efforts of trying and experiments, we propose two feasible solution methods for these two problems. Then, we combine these solution methods with the method for regularizing and computing discrete cepstrum coefficients to form a spectral-envelope estimation scheme. This scheme has been verified, by measuring spectral-envelope approximation error, to be much better than the original scheme. In addition, we have applied this scheme to build a voice-timbre transformation system. This system demonstrates that the proposed estimation scheme is very effective.