Reconocimiento de emociones a partir de voz basado en un modelo emocional continuo

In this thesis, we worked on emotion recognition from the speech signal. To address this problem, we have adopted a psychological continuous emotions model. With this emotion model, emotions are studied in a broader sense of what has been done traditionally. It has been thoroughly explored the acous...

Descripción completa

Detalles Bibliográficos
Autor: HUMBERTO PEREZ ESPINOSA
Tipo de recurso: tesis doctoral
Estado:Versión aceptada para publicación
Fecha de publicación:2013
País:México
Institución:Instituto Nacional de Astrofísica, Óptica y Electrónica
Repositorio:Repositorio Institucional del INAOE
Idioma:español
OAI Identifier:oai:inaoe.repositorioinstitucional.mx:1009/795
Acceso en línea:http://inaoe.repositorioinstitucional.mx/jspui/handle/1009/795
Access Level:acceso abierto
Palabra clave:info:eu-repo/classification/Reconocimiento de patrones/Pattern recognition
info:eu-repo/classification/Clasificación de patrones/Pattern classification
info:eu-repo/classification/Estructuras de datos/Data structures
info:eu-repo/classification/Procesos de Markov/Markov processes
info:eu-repo/classification/cti/1
info:eu-repo/classification/cti/12
info:eu-repo/classification/cti/1203
Descripción
Sumario:In this thesis, we worked on emotion recognition from the speech signal. To address this problem, we have adopted a psychological continuous emotions model. With this emotion model, emotions are studied in a broader sense of what has been done traditionally. It has been thoroughly explored the acoustic feature space on realistic scenarios. We applied soft computing and probabilistic techniques for estimating emotional states. This work contributes to the understanding of speech elements that help to determine the emotions and to create a pattern recognition method based on a continuous emotional model suitable for spontaneous emotions detection. We experimented with various types of acoustic and linguistic features, including new features used in other fields. We had used feature selection techniques to find the most important ones. In addition, we have studied the importance of these attributes on multilingual data, we propose a method for representing emotional states based on fuzzy clustering and studied the prediction of emotional states based on context. The results obtained in the estimation of emotion primitives and classification of basic emotions using our feature set and methods are comparable with the best results in the state of the art.