Speech parameter generation considering LSP ordering property for HMM-based speech synthesis

LSP has many advantages for speech representation, especially correlates well to spectrum formants as long as the LSP parameters are strictly ordered and bounded. This ordering property cannot be guaranteed during HMM-based speech synthesis when LSP is adopted as the spectrum feature, because diagon...

Full description

Saved in:
Bibliographic Details
Main Authors: Shijun Qian, Huanliang Wang, Wenjiang Pei, Ping Zou, Kai Wang
Format: Conference Proceeding
Language:English
Subjects:
Equations
hidden Markov model
Hidden Markov models
line spectral pair
Mathematical model
ordering property
parameter generation
Speech
Speech synthesis
Training
Vectors
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:LSP has many advantages for speech representation, especially correlates well to spectrum formants as long as the LSP parameters are strictly ordered and bounded. This ordering property cannot be guaranteed during HMM-based speech synthesis when LSP is adopted as the spectrum feature, because diagonal covariance is utilized and correlation between LSP dimensions is ignored, with the result that unstable issue will be caused in synthesized speech. In this paper, we modify the parameter generation criterion to preserve ordering property of generated LSPs, by considering not only the likelihoods for HMM and GV maximized in conventional method but also a mis-orderings penalty. Experimental results show that the proposed method can alleviate the mis-orderings significantly and achieve high quality synthesizing performance when the penalty weight is selected appropriately.
ISSN:2219-5491
2219-5491
DOI:10.1109/LSP.2012.2203593