Conjuntos difusos tipo 2 aplicados a la comparación difusa de patrones para clasificación de llanto de infantes con riesgo neurológico
The interest for analyze the baby’s cry has increased in the last years, nowadays computational models are in use to automate the identification of pathologies in the baby’s cry. For the classification of baby’s cry have been used methods based on artificial neural networks. In this thesis we propos...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis de maestría |
| Estado: | Versión aceptada para publicación |
| Fecha de publicación: | 2009 |
| País: | México |
| Institución: | Instituto Nacional de Astrofísica, Óptica y Electrónica |
| Repositorio: | Repositorio Institucional del INAOE |
| Idioma: | español |
| OAI Identifier: | oai:inaoe.repositorioinstitucional.mx:1009/454 |
| Acceso en línea: | http://inaoe.repositorioinstitucional.mx/jspui/handle/1009/454 |
| Access Level: | acceso abierto |
| Palabra clave: | info:eu-repo/classification/Teoría de conjuntos difusos/Fuzzy set theory info:eu-repo/classification/Clasificación de patrones/Pattern classification info:eu-repo/classification/Emparejamiento de patrones difusos/Fuzzy pattern matching info:eu-repo/classification/Análisis del llanto infantil/Infant cry analysis info:eu-repo/classification/cti/1 info:eu-repo/classification/cti/12 info:eu-repo/classification/cti/1203 |
| Sumario: | The interest for analyze the baby’s cry has increased in the last years, nowadays computational models are in use to automate the identification of pathologies in the baby’s cry. For the classification of baby’s cry have been used methods based on artificial neural networks. In this thesis we propose a classification method based on the fuzzy pattern matching by applying type-2 fuzzy sets. This method use the Information of a element about its membership to a particular class and its relative importance with respect to other classes. We tested the classification with samples of baby´s crying labeled as asphyxia, hyperbilirubinemia and normal. Class hyperbilirubinaemia has not been classified automatically before. The classification process includes the extraction of characteristics of the signal crying, which are used to train the method proposed in this thesis, and subsequently to realize tests. To extract the characteristics of signal crying we used four methods commonly used for the analysis of speech: LPC (Linear Predictive Coding), MFCC (Mel Frequency Cepstral Coefficients), Intensity and Cochleograms. We performed tests with different combinations of characteristics and classes. The best results for classification using the three clases (hyperbilirubinemia, asphyxia and normal) were obtained with the combination of the characteristics LPC and Cochleograms, where we obtained an accuracy of 91.74 %. In the classification using two classes (hiperbilirrubinemia and normal) the best result was obtained with the combination of the characteristics LPC and MFCC with an accuracy of 95.56%. And with the combination: asphyxia and normal, better results were obtained (99.02% of accuracy) than similar works of classification of baby’s cry previously reported. |
|---|