Publications

Author: evilches
October 14, 2009

Tittle: Data Mining Applied to Acoustic Bird Species Recognition

Authors: Erika Vilches, Ivan A. Escobar, Edgar E. Vallejo and Charles E. Taylor

Abstract: In this work we explore the application of data mining techniques to the problem of acoustic recognition of bird species. Most bird song analysis tools produce a large amount of spectral and temporal attributes from the acoustic signal. The identification of distinctive features has become critical in resource constrained applications such as habitat monitoring by sensor networks. Reducing computational requirements makes affordable to run a classifier on devices with power consumption constraints, such as nodes in a sensor network. Experimental results demonstrate that considerable dimensionality reduction can be achieved without significant loss in classification efficiency.

Conference: 18th International Conference of Pattern Recognition 2006 (ICPR 06), August 20-24, 2006, Hong Kong

Publisher: IEEE

Year: 2006

Download link: Data Mining Applied to Acoustic Bird Species Recognition

__________________________________________________________________________________________________________________

Thesis to opt for the degree of Master of Science in Computer Science, Artificial Intelligence specialty

Tittle: Application and Comparison of Artificial Intelligence and Statistics Methodologies for Acoustic Bird Species Recognition

Author: Erika Vilches González

Supervisor: Dr. Edgar Emmanuel Vallejo Clemente

External Supervisor: Dr. Charles Taylor

Abstract: In this work, we explore the application of Artificial Intelligence and Statistics techniques to the problem of acoustic recognition of bird species based on their song production and contrast these results with the traditional approaches based on Hidden Markov Models and Neural Networks. Previous work has shown that large collections of spectral and temporal attributes are needed in order to represent the structure of bird songs. First, with the data mining approach, we reduced the features extracted from the bird songs and used its classification techniques to discriminate among bird species. More specifically, J4.8, a decision tree algorithm was implemented in order to extract the most significant attributes. Naive-Bayes, a probabilistic classifier, was applied over the resulting J4.8 reduced attribute data set. This was done in order to compare the efficiency of the reduced attribute data results against those obtained from the complete attribute data set. Next, the association rules algorithm was applied to the reduced attribute data set in order to find natural relationships between the attributes. Finally the efficiency of data mining classification algorithms was compared against those of HMM’s and Neural Networks, proving that the data mining results yielded higher accuracy and simplicity.

Keywords: Data mining, Hidden Markov Models, Neural Networks, Bird Song, Species Recognition, Feature Extraction.

Number of Pages: 85

Date: August 2006

Download link: Application and Comparison of Artificial Intelligence and Statistics Methodologies for Acoustic Bird Species Recognition

__________________________________________________________________________________________________________________

Tittle: Targeting Input Data for Acoustic Bird Species Recognition Using Data Mining and HMMs

Authors: Erika Vilches, Ivan A. Escobar, Edgar E. Vallejo and Charles E. Taylor

Abstract: In this paper we propose the integration of Data Mining with Hidden Markov Models when applied to the problem of acoustic bird species recognition. We ?rst show how each of them is applied on an individual manner, contrast their results and devise a model to combine them for targeted classi?cations. Previous work has shown that large collections of spectral attributes are needed in order to represent the structure of bird songs, therefore elevating the computational requirements when applied to distributed sensor networks. Data Mining is used to reduce the dimensionality of the spectral attributes and for classi?cation. Hidden Markov models represent a traditional approach and require strong song preprocessing. Our results show that Data Mining can yield ef?cient results with low requirements and could serve to target HMMs input parameters.

Conference: Seventh IEEE International Conference on Data Mining Workshops, October 20-31 2007, Omaha, Nebraska

Publisher: IEEE

Year: 2007

Download link: Targeting Input Data for Acoustic Bird Species Recognition Using Data Mining and HMMs

__________________________________________________________________________________________________________________

Tittle: Self-organizing acoustic categories in sensor arrays

Authors: Ivan A. Escobar, Erika Vilches, Edgar E. Vallejo, Martin L. Cody and Charles E. Taylor

Abstract: In this paper, we explore the emergence of acoustic categories in sensor arrays. We describe a series experiments on the automatic categorization of species and individual birds using self-organizing maps. Experimental results showed that meaningful acoustic categories can arise as self-organizing processes in sensor arrays. In addition, we discuss how distributed categorization could be used for the emergence of symbolic communication in these platforms.

Conference: 9th European Conference on Artificial Life, September 10-14, 2007 Lisbon, Portugal

Publisher: Springer

Year: 2007

Download link: Self-organizing acoustic categories in sensor arrays

__________________________________________________________________________________________________________________

Tittle: Sistemas de Ficheros EXT2, HFS+ y NTFS

Authors: Ivan A. Escobar y Erika Vilches

Abstract: En el mundo en el que vivimos hoy en día, resulta muy práctico que utilicemos los ordenadores para todo tipo de actividades. En la ofi cina nos ponen un ordenador inmenso que debemos saber manejar como expertos, y en algunos casos nos dotan con equipos portátiles para que nos llevemos el trabajo sin terminar a casa o podamos llevarlo a las reuniones de trabajo. En casa es común tener un ordenador para explorar la red, para contactar a los miembros lejanos de la familia y para resolver las necesidades personales.

Journal/Publisher: Linux+, Julio 2007

Year: 2007

Download link: Sistemas de ficheros EXT2, HFS+ y NTFS

__________________________________________________________________________________________________________________

Tittle: Minería de Datos

Authors: Erika Vilches e Ivan A. Escobar

Abstract: En la actualidad, muchas de las decisiones importantes que se toman alrededor del mundo se basan en observaciones y/o eventos que han sido previamente registrados de alguna forma en una base o modelo de datos. Esta información puede llevar a analistas de mercado a tomar decisiones en cuanto a la compra o venta de acciones, a médicos que trabajan en clínicas de fertilización a decidir entre diferentes muestras de embriones cuyas sesenta propiedades pueden determinar cual sería la mejor opción para transferirlos al útero de una mujer y tener probabilidades exitosas de embarazo y a granjeros a decidir cuales son sus muestras de ganado más útiles a preservar antes de la próxima venta o trueque.

Journal/Publisher: Linux+, September 2007

Year: 2007

Download link: Minería de Datos

__________________________________________________________________________________________________________________

Tittle: Producción de Templates

Authors: Erika Vilches e Ivan A. Escobar

Abstract: Pocas personas saben cómo crear sus propios templates a partir de un diseño en Photoshop que les agrade y cómo utilizar Dreamweaver para automatizar la propagación de cambios en un template a todos los archivos producidos a partir del mismo. La automatización que este proceso genera puede ahorrarnos mucho tiempo en la edición de sitios web y aumentar nuestra productividad.

Journal/Publisher: .psd Photoshop Solutions for Designers, January 2008

Year: 2008

Download link: Produccion de templates