September 22-25, 1997
Linear Discriminant Analysis (LDA) has been widely applied to speech recognition resulting in improved recognition performance and improved robustness. LDA designs a linear transformation that projects a m-dimensional space on a m-dimensional space (m < n) such that the class separability is maximum. This paper presents new results related to our previous work  on nonlinear discriminant analysis (NLDA) based on the discriminant properties of Artificial Neural Networks (ANN) and more particularly MLP. Experiments performed on the isolated word large vocabulary Phone- book database show that NLDA provides a method for designing discriminant features particularly efficient as well for continuous densities HMM as for hybrid HMM/ANN recognizers.
Bibliographic reference. Fontaine, Vincent / Ris, Christophe / Boite, Jean-Marc (1997): "Nonlinear discriminant analysis for improved speech recognition", In EUROSPEECH-1997, 2071-2074.