International Symposium on Chinese Spoken Language Processing
August 23-24, 2002
Constrained Maximum A Posteriori Approach for Speech Enhancement
Chuan Jia, Jian Zhang, Bo Xu
National Laboratory of Pattern Recognition,
Institute of Automation,
Chinese Academy of Sciences, Beijing, China
The maximum a posteriori estimator based on HMMís is
successful to some degree because of the incorporation of prior
knowledge of speech and markovian properties of the models.
The enhanced speech quality is, however, not satisfying at low
input SNR. In order to improve speech quality at low input SNR,
this paper proposes a method that incorporates codebook
constrained Wiener filter into MAP framework to impose
spectral constraints on estimated speech signals. The objective
measures, global SNR and Itakura-Saito distortion measure,
verified the quality improvement of the proposed method.
Jia, Chuan / Zhang, Jian / Xu, Bo (2002):
"Constrained maximum a posteriori approach for speech enhancement",
In ISCSLP 2002, paper 97.