Powered by OpenAIRE graph
Found an issue? Give us feedback
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/ LAReferencia - Red F...arrow_drop_down
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
versions View all 1 versions
addClaim

Modelización de secuencias para el reconocimiento automático de patrones

Authors: Martínez, César Ernesto;

Modelización de secuencias para el reconocimiento automático de patrones

Abstract

La modelización de secuencias es un problema de gran interés en el reconocimiento de patrones. En el mismo se busca diseñar y construir sistemas especializados en capturar las particularidades de tramos distintivos de las secuencias y su estructura de repetición. Estos sistemas pueden ser empleados tanto para la clasificación (modelos discriminativos) como para la síntesis (modelos generativos). En este trabajo se avanza sobre la extracción de características y modelización en dos dominios de aplicación: las clasificación de imágenes de cromosomas y el reconocimiento robusto de señales de habla. En el primero, se proponen nuevas parametrizaciones que explotan la variabilidad de las bandas de grises a lo largo de los cromosomas. Se introducen nuevas formas de clasificar estos patrones basadas en redes neuronales recurrentes y modelos ocultos de Markov continuos. Se propone, además, un algoritmo de post-clasificación contextual que clasifica todos los cromosomas de una célula en su conjunto. En la representación de la señal de la voz, se proponen nuevas alternativas bioinspiradas de parametrización, que modelan la activación de la corteza auditiva primaria en respuesta a los estímulos sonoros. Los patrones generados mediante una representación rala se aplican a las tareas de clasificación robusta de fonemas y limpieza de señales de habla inmersas en ruido. Los resultados obtenidos muestran que estas técnicas logran extraer las pistas útiles para el reconocimiento y/o recuperar la información que preserva objetivamente la calidad de las señales limpiadas, mostrando ventajas en el desempeño respecto a los métodos clásicos.

Sequence modeling is a problem of interest in the field of pattern recognition. It is aimed at the design and building of specialized systems that capture the particularities of distinctive segments of the sequences and their repetition structure. These systems could be used for both classification (discriminative models) as well as for synthesis (generative models). In this work we present progresses on the feature extraction and sequence modeling in two domains of applications: classification of chromosome images and robust speech recognition. In the first one, new parameterizations are here proposed to exploit the variability of the gray bands along the chromosomes. New ways to classify these patterns based on recurrent neural networks and continuous hidden Markov models are introduced. Furthermore, a contextual post-classification algorithm is proposed to carry out a relocation of chromosomes in each class according to the expected number of chromosomes in a cell. For the speech signal representation, new bioinspired alternatives to the speech parameterization are proposed, which model the activation of the primary auditory cortex in response to sound stimuli. The sparse coding patterns obtained are applied to robust phoneme classification and speech denoising. The obtained results show that these techniques can extract useful clues for recognition and retrieval of information that objetively preserve the quality of the denoised signals, with performance benefits over other methods previously reported.

Fil: Martínez, César Ernesto. Universidad Nacional del Litoral. Facultad de Ingeniería y Ciencias Hídricas; Argentina.

Agencia Nacional de Promoción Científica y Tecnológica

Universidad Nacional de Entre Ríos

Universidad Nacional del Litoral

Country
Argentina
Related Organizations
Keywords

Automatic karyotyping, Sparse coding, Pattern recognition, Robust speech recognition, Clasificación de cromosomas, Chromosome classification, Reconocimiento robusto del habla, Cariotipado automático, Representación cortical aproximada, Approximated cortical representation, Representaciones ralas, Reconocimiento de patrones

  • BIP!
    Impact byBIP!
    selected citations
    These citations are derived from selected sources.
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
Powered by OpenAIRE graph
Found an issue? Give us feedback
selected citations
These citations are derived from selected sources.
This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Citations provided by BIP!
popularity
This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
BIP!Popularity provided by BIP!
influence
This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Influence provided by BIP!
impulse
This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
BIP!Impulse provided by BIP!
0
Average
Average
Average
Green