A flexible temporal decomposition-based spectral modification method using asymmetric Gaussian mixture model (音声)

概要

論文の詳細を見る
This paper presents a new spectral modification method to deal with two drawbacks of standard spectral modification methods, insufficient smoothness of the modified spectra between frames and ineffective spectral modification. A speech analysis technique called Temporal Decomposition (TD), which decomposes speech into event targets and event functions, is used to effectively model the spectral evolution. Instead of modifying the speech spectra frame by frame, we only need to modify event targets and event functions. We then proposed a new method to model and modify an event target by using asymmetric Gaussian mixture model (AGMM). Experimental results show that the effectiveness of the proposed spectral modification method is verified in terms of the smoothness of modified speech and effective spectral modification.
社団法人電子情報通信学会の論文
2007-07-19