Evaluation of Combinational Use of Discriminant Analysis-Based Acoustic Feature Transformation and Discriminative Training

Makoto SAKAI  Norihide KITAOKA  Yuya HATTORI  Seiichi NAKAGAWA  Kazuya TAKEDA  

Publication
IEICE TRANSACTIONS on Information and Systems   Vol.E93-D   No.2   pp.395-398
Publication Date: 2010/02/01
Online ISSN: 1745-1361
DOI: 10.1587/transinf.E93.D.395
Print ISSN: 0916-8532
Type of Manuscript: LETTER
Category: Speech and Hearing
Keyword: 
speech recognition,  feature extraction,  discriminative training,  

Full Text: PDF(78.2KB)>>
Buy this Article




Summary: 
To improve speech recognition performance, acoustic feature transformation based on discriminant analysis has been widely used. For the same purpose, discriminative training of HMMs has also been used. In this letter we investigate the effectiveness of these two techniques and their combination. We also investigate the robustness of matched and mismatched noise conditions between training and evaluation environments.