On the determination of epsilon during discriminative GMM training

• Autor(a) principal: Guido, Rodrigo Capobianco [UNESP]
• Data de Publicação: 2010
• Outros Autores: Chen, Shi-Huang [UNESP], Junior, Sylvio Barbon [UNESP], Souza, Leonardo Mendes [UNESP], Vieira, Lucimar Sasso [UNESP], Rodrigues, Luciene Cavalcanti [UNESP], Escola, Joao Paulo Lemos [UNESP], Zulato, Paulo Ricardo Franchi [UNESP], Lacerda, Michel Alves [UNESP], Ribeiro, Jussara [UNESP]
• Tipo de documento: Artigo de conferência
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.1109/ISM.2010.66; http://hdl.handle.net/11449/72054
• topic: Discriminative training of Gaussian Mixture Models (GMMs); Markov Models; Speaker identification; Speech recognition; Discriminative training; Gaussian mixture models; Gradient descent algorithms; Gradient Descent method; Iteration step; Newton-Raphson iterative method; Second orders; Speaker recognition; Gaussian distribution; Iterative methods; Loudspeakers; Markov processes

Discriminative training of Gaussian Mixture Models (GMMs) for speech or speaker recognition purposes is usually based on the gradient descent method, in which the iteration step-size, ε, uses to be defined experimentally. In this letter, we derive an equation to adaptively determine ε, by showing that the second-order Newton-Raphson iterative method to find roots of equations is equivalent to the gradient descent algorithm. © 2010 IEEE.