Title :
Parameter Generation Considering LSP Ordering Property for HMM-Based Speech Synthesis
Author :
Qian, Shijun ; Wang, Huanliang ; Pei, Wenjiang ; Wang, Kai
Author_Institution :
Sch. of Inf. Sci. & Eng., Southeast Univ., Nanjing, China
Abstract :
The ordering property of LSP cannot be guaranteed during HMM-based speech synthesis when LSP is adopted as the spectrum feature, with the result that naturalness of synthetic speech will degrade. This letter introduces a modified parameter generation method to preserve ordering property of generated LSPs, by not only maximizing the likelihoods of HMM and global variance (GV) as in conventional method but also minizing a penalty on mis-orderings. 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.
Keywords :
hidden Markov models; speech synthesis; HMM likelihoods; HMM-based speech synthesis; LSP ordering property; global variance; hidden Markov model; high quality synthesizing performance; misorderings; modified parameter generation method; spectrum feature; synthetic speech; Acoustics; Equations; Hidden Markov models; Mathematical model; Speech; Speech synthesis; Training; Speech synthesis; hidden Markov model; line spectral pair; parameter generation;
Journal_Title :
Signal Processing Letters, IEEE
Conference_Location :
6/6/2012 12:00:00 AM
DOI :
10.1109/LSP.2012.2203593