ISCA Archive SPKD 2008
ISCA Archive SPKD 2008

A unified approach for F0 extraction and aperiodicity estimation based on a temporally stable power spectral representation

Hideki Kawahara, Masanori Morise, Toru Takahashi, Ryuichi Nisimura, Hideki Banno, Toshio Irino

A power spectrum estimation method for periodic signals was proposed to provide temporally stable representation and has been applied to reformulate STRAIGHT, a system for speech analysis modification and synthesis based on stable spectral envelope estimation. This article proposes a specialized F0 detector based on a ratio between this stable spectrum and corresponding spectral envelope. By allocating multiple specialized F0 detectors and integrating individual clues, the proposed method selectively detects only fundamental components and yields a probability measure for each estimate. It also provides a method to estimate aperiodicity in each frequency band by making use of estimated fundamental frequency information to design a quadrature signal on the frequency axis for filtering periodic spectral component due to the signal periodicity. The proposed method shed new lights on source filter representation/decomposition of speech signals.


Cite as: Kawahara, H., Morise, M., Takahashi, T., Nisimura, R., Banno, H., Irino, T. (2008) A unified approach for F0 extraction and aperiodicity estimation based on a temporally stable power spectral representation. Proc. ISCA ITRW on Speech Analysis and Processing for Knowledge Discovery, paper 043

@inproceedings{kawahara08_spkd,
  author={Hideki Kawahara and Masanori Morise and Toru Takahashi and Ryuichi Nisimura and Hideki Banno and Toshio Irino},
  title={{A unified approach for F0 extraction and aperiodicity estimation based on a temporally stable power spectral representation}},
  year=2008,
  booktitle={Proc. ISCA ITRW on Speech Analysis and Processing for Knowledge Discovery},
  pages={paper 043}
}