Minimum classification error learning for sequential data in the wavelet domain

  • Authors:

  • D. Tomassi;D. H. Milone;L. Forzani

  • Affiliations:

  • Centro de Investigación en Señales e Inteligencia Computacional, Facultad de Ingeniería y Ciencias Hídricas, Universidad Nacional del Litoral-CONICET, Ciudad Universitaria CC 2 ...;Centro de Investigación en Señales e Inteligencia Computacional, Facultad de Ingeniería y Ciencias Hídricas, Universidad Nacional del Litoral-CONICET, Ciudad Universitaria CC 2 ...;Departamento de Matemática, Facultad de Ingeniería Química, Universidad Nacional del Litoral, Argentina and Instituto de Matemática Aplicada del Litoral-CONICET, Argentina

  • Venue:
  • Pattern Recognition
  • Year:
  • 2010

Quantified Score

Hi-index 0.01

Visualization

Abstract

Wavelet analysis has found widespread use in signal processing and many classification tasks. Nevertheless, its use in dynamic pattern recognition have been much more restricted since most of wavelet models cannot handle variable length sequences properly. Recently, composite hidden Markov models which observe structured data in the wavelet domain were proposed to deal with this kind of sequences. In these models, hidden Markov trees account for local dynamics in a multiresolution framework, while standard hidden Markov models capture longer correlations in time. Despite these models have shown promising results in simple applications, only generative approaches have been used so far for parameter estimation. The goal of this work is to take a step forward in the development of dynamic pattern recognizers using wavelet features by introducing a new discriminative training method for this Markov models. The learning strategy relies on the minimum classification error approach and provides re-estimation formulas for fully non-tied models. Numerical experiments on phoneme recognition show important improvement over the recognition rate achieved by the same models trained using maximum likelihood estimation.