Reconocimiento automático de voz emotiva con memorias asociativas Alfa–Beta SVM

 

Automatic Emotional Speech Recognition with Alpha–Beta SVM Associative Memories

 

José Francisco Solís Villarreal*, Cornelio Yáñez Márquez** y Sergio Suárez Guerra***

 

Centro de Investigación en Computación del Instituto Politécnico Nacional, México, D.F. (*tlilectic.mixtzin@gmail.com, **cyanez@cic.ipn.mx, ***ssuarez@cic.ipn.mx).

 

Manuscrito recibido el 03 de marzo de 2011.
Manuscrito aceptado para su publicación el 22 de junio de 2011.

 

Resumen

Una de las de investigación de mayor interés y con más crecimiento en la actualidad, dentro del área de procesamiento de voz, es el reconocimiento automático de emociones, el cual consta de 2 etapas; la primera es la extracción de parámetros a partir de la señal de voz y la segunda es la elección del modelo para hacer la tarea de clasificación. La problemática que actualmente existe es que no se han identificado aún los parámetros más representativos del problema ni tampoco se ha encontrado al mejor clasificador para hacer la tarea. En este artículo se introduce un nuevo modelo asociativo de reconocimiento automático de voz emotiva basado en las máquinas asociativas Alfa–Beta SVM, cuyas entradas se han codificado como representaciones bidimensionales de la energía de las señales de voz. Los resultados experimentales muestran que este modelo es competitivo en la tarea de clasificación automática de emociones a partir de señales de voz [1].

Palabras Clave: Reconocimiento de voz emotiva, memorias asociativas Alfa–Beta SVM, procesamiento de voz.

 

Abstract

One of the research lines of interest and more growth at present, within the area of voice processing is automatic emotion recognition. It is vitally important the study of speech signal not only to extract information about what is being said, but how is being said, this in order to be closer to the human–machine interaction. In literature the procedure of automatic emotion recognition consists of two stager, the first is the extraction of parameters from the voice signal and the second is the choice of model for the classification task, the problem that currently exists is not yet identified the most representative parameters of the problem nor has found the best classifier for the task, but have not yet been tested several models, this paper presents a two–dimensional representation of energy as data entry for Alpha–Beta associative machines SVM (Support Vector Machine) for the classification of emotions.

Key words: Emotional speech recognition, Alpha–Beta SVM associative memories, voice processing.

 

DESCARGAR ARTÍCULO EN FORMATO PDF

 

AGRADECIMIENTOS

Los autores agradecen el apoyo de las siguientes instituciones para la realización de esta obra: Secretaría de Investigación y Posgrado, Secretaría Académica, COFAA y CIC del Instituto Politécnico Nacional, CONACyT y Sistema Nacional de Investigadores (SNI); específicamente, los proyectos SIP–20090807, SIP–20101709 y SIP–20110661.

 

REFERENCIAS

[1] Berlin emotional speech database, http://www.expresive–speech.net/.         [ Links ]

[1] T. Vogt, E. André, and J. Wagner, Automatic Recognition of Emotions from Speech: A Review of the Literature and Recommendations for Practical Realization, Affect and Emotion in Human–Computer Interaction: From Theory to Applications, Springer–Verlag, Berlin, Heidelberg, 2008.         [ Links ]

[2] J. Cichosz and K. Slot, "Emotion recognition in speech signal using emotion–extracting binary decision trees," in Proceedings of Affective Computing and Intelligent Interaction, Lisbon, Portugal, 2007.         [ Links ]

[3] T. Vogt, and E. André, "Improving automatic emotion recognition from speech via gender differentiation," in Proceedings of Language Resources and Evaluation Conference, 2006.         [ Links ]

[4] Z. Xiao, E. Dellandrea, W. Dou, and L. Chen, "Hierarchical classification of emotional speech," IEEE Transactions on Multimedia, 2007.         [ Links ]

[5] H. Ian, and F. Eibe, Data Mining: Practical machine learning tools and techniques, 2nd Edition, Morgan Kaufmann, San Francisco, available online at http://www.cs.waikato.ac.nz/ml/weka/, 2005.         [ Links ]

[6] D. Ververidis and C Kotropoulos, "A state of the art review on emotional speech databases," in Proceedings of 1st Richmedia Conference, 2003, pp. 109–119.         [ Links ]

[7] L. López–Leyva, C. Yáñez–Márquez, R. Flores–Carapia, and O. Camacho–Nieto, "Handwritten Digit Classification Based on Alpha–Beta Associative Model," in Progress in Pattern Recognition, Image Analysis and Applications. LNCS 5197, Proc. 13th Iberoamerican Congress on Pattern Recognition CIARP 2008, Havana, Cuba, 2008.         [ Links ]

[8] L. López–Leyva, C. Yáñez–Márquez, and I. López–Yáñez, "A new efficient model of support vector machines: ALFA–BETA SVM," in 23rd ISPE International Conference on CAD/CAM, Robotics and Factories of the Future, Bogotá, Colombia, 2007.         [ Links ]

[9] C. Yáñez–Márquez, Associative Memories Based on Order Relations and Binary Operators (In Spanish). PhD Thesis. Center for Computing Research, Mexico, 2002.         [ Links ]

[10] T. Kohonen, Self–Organization and Associative Memory, Springer–Verlag, Berlin Heidelberg New York, 1989.         [ Links ]

[11] M. H. Hassoun, Associative Neural Memories, Oxford University Press, New York, 1993.         [ Links ]

[12] T. Kohonen, "Correlation Matrix Memories," IEEE Transactions on Computers, 21(4), 353–359, 1972.         [ Links ]

[13] M. El Ayadi, M. Kamel, and F. Darray, "Survey on speech emotion recognition: Features, classification schemes and databases," Pattern Recognition, vol. 44, March, (2011) 572–587.         [ Links ]

[14] S. Zhang, L. Li, and Z. Zhao, "Spoken emotion recognition using kernel discriminant locally linear embedding," Electronics Letters, vol 46, 1344–1346, 2010.         [ Links ]