Eigenspace-based Linear Transformation Approach for Rapid Speaker Adaptation

This paper presents our recent effort on the development of the eigenspace-based linear transformation approach for rapid speaker adaptation. The proposed approach toward prior density selection for the MAPLR framework was developed by introducing a priori knowledge analysis on the training speakers via probabilistic principal component analysis (PPCA), so as to construct an eigenspace for speaker-specific full regression matrices as well as to derive a set of bases called eigen-transformations. The prior densities of MAPLR transformations for each outside speaker are then chosen in the space spanned by the first few eigen-transformations. By incorporating the PPCA model of transformation parameters into the MAPLR scheme, the number of free parameters can be significantly reduced, while the underlying structure of the acoustic space as well as the precise modeling of the inter-dimensional correlation among the model parameters can be well preserved. Rapid supervised adaptation experiments showed that the proposed approach not only is superior to the conventional MLLR approach using either diagonal or block-diagonal regression matrices, but also outperformed by a great amount the full-matrix MLLR with either a global transformation or multiple transformations corresponding to different phonetic classes.