Publication: Análisis y clasificación de emociones en personas con discapacidad visual usando señales EEG y técnicas aprendizaje automático
Loading...
Identifiers
Publication date
2021
Defense date
2022-02-24
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
Actualmente, el reconocimiento y clasificación de las emociones humanas a partir del análisis de la actividad cerebral de una persona, sigue representando un desafío y un problema a resolver, debido a que la percepción de las emociones entre las personas difiere y a la dificultad para generalizar el conjunto de señales cerebrales. Para abordar este problema es necesario integrar diferentes disciplinas que en conjunto permitan dar una solución a este problema.
La computación afectiva es una rama de la inteligencia artificial, que propone integrar en las computadoras la capacidad de reconocer las emociones de una persona para mejorar la interacción entre computadoras y personas y su experiencia en el uso de la tecnología. Por otro lado, la interfaz cerebro-computadora es una tecnología utilizada para registrar y analizar la actividad cerebral. Por lo tanto, se ha utilizado para estudiar las emociones humanas a partir de un conjunto de señales cerebrales.
Por otro lado, trabajos de investigación previos han propuesto estudiar el comportamiento emocional de una persona para clasificar sus emociones, a través de diferentes fuentes de información como expresiones faciales, tono de voz y señales cerebrales, entre otras. Sin embargo, estos estudios no están disponibles en todas las situaciones y no han considerado a todas las personas y sus discapacidades.
En consecuencia, esta investigación presenta el diseño de un marco para el reconocimiento de emociones dirigido a personas con discapacidad visual. En primer lugar, el marco se adapta a este tipo de personas y su discapacidad. En segundo lugar, se utilizan estímulos auditivos de corta duración para inducir estados afectivos, y se ha definido un componente de adquisición y extracción de características de las señales cerebrales. Finalmente, se han utilizado técnicas de análisis para modelar las señales cerebrales y se han evaluado modelos de aprendizaje automático para clasificar las emociones. El resultado esperado es obtener nueva evidencia sobre la aplicación de una interfaz cerebro-computadora y computación afectiva en un escenario no evaluado previamente, el reconocimiento de información de los estados afectivos de personas con discapacidad visual y la clasificación de sus emociones.
La metodología definida para lograr el objetivo de clasificar las emociones en personas con discapacidad visual y las actividades para el desarrollo de la tesis doctoral son las siguientes:
● Estado del arte. En esta fase se realiza un análisis de trabajos relacionados sobre interfaz cerebro-computadora y computación afectiva dirigidos a personas con discapacidad visual y la clasificación de sus estados afectivos.
● Solución propuesta. En esta etapa se presenta el framework para la clasificación de las emociones en un escenario orientado a personas con discapacidad visual. Asimismo, se describen los elementos y componentes del framework de cada una de sus fases.
● Evaluación y validación. En esta sección se presentan las actividades que permiten la evaluación y validación de las fases del framework. Además, se describen enfoques emocionales basados en señales cerebrales para identificar qué emociones muestran comportamientos diferentes en relación con las señales cerebrales.
● Documentación y conclusiones. En esta sección se documenta cada aspecto relativo a la tesis doctoral y se presentan a manera de conclusión los resultados que surgen de la investigación.
Según los resultados, el framework puede clasificar las emociones positivas y negativas, pero la experimentación realizada también muestra que el rendimiento del marco depende del número de características en el conjunto de datos y la calidad de las señales de EEG es un factor determinante.
Currently, the recognition and classification of human emotions from the analysis of a person's brain activity still represent a challenge and a problem to be solved, due to the fact that the perception of emotions among people differs and the difficulty to generalize the set of brain signals. For this, it is necessary to integrate different disciplines that allow a solution to this problem. Affective computing is a branch of artificial intelligence, which proposes integrating into computers the ability to recognize a person's emotions to improve the interaction between computers and people and, their experience in the use of technology. On the other hand, the brain-computer interface is a technology used to record and analyze brain activity. Therefore, it has been used to study human emotions from a set of brain signals. On the other hand, previous research works have proposed studying the emotional behavior of a person to classify their emotions, through different sources of information such as facial expressions, voice tone, and brain signals, among others. However, these studies are not available in all situations and have not considered all people and their disabilities. Consequently, this research presents the design of a framework for the recognition of emotions aimed at people with visual disabilities. Firstly, the framework is adapted to this type of people and their disability. Second, short-duration auditory stimuli are used to induce affective states, and a component of acquisition and features extraction of brain signals has been defined. Finally, analysis techniques have been used to model the brain signals, and machine learning models have been evaluated to classify emotions. The expected result is to obtain new evidence on applying a brain-computer interface and affective computing in a scenario not previously evaluated, the recognition of information of the affective states of people with visual disabilities, and the classification of their emotions. The methodology defined to achieve the objective of classifying emotions in people with visual disabilities and the activities for the development of the doctoral thesis are the following: ● State of the art. In this phase, an analysis of related works is carried out on brain-computer interface and affective computing aimed at people with visual disabilities and the classification of their affective states. ● Proposed solution. In this stage, the framework for the classification of emotions in a scenario oriented to people with visual disabilities is presented. Likewise, the elements and components of the framework of each of its phases are described. ● Evaluation and validation. In this section, the activities that allow the evaluation and validation of the framework phases are presented. Additionally, brain signal-based emotion approaches are described to identify which emotions show different behavior in relation to brain signals. ● Documentation and conclusions. In this section, each aspect related to the doctoral thesis is documented and the results that emerge from the research are presented as a conclusion. Based on the results, the framework can classify positive and negative emotions, but the experimentation conducted also shows that the framework's performance depends on the number of features in the dataset and the quality of the brain signals is a determining factor.
Currently, the recognition and classification of human emotions from the analysis of a person's brain activity still represent a challenge and a problem to be solved, due to the fact that the perception of emotions among people differs and the difficulty to generalize the set of brain signals. For this, it is necessary to integrate different disciplines that allow a solution to this problem. Affective computing is a branch of artificial intelligence, which proposes integrating into computers the ability to recognize a person's emotions to improve the interaction between computers and people and, their experience in the use of technology. On the other hand, the brain-computer interface is a technology used to record and analyze brain activity. Therefore, it has been used to study human emotions from a set of brain signals. On the other hand, previous research works have proposed studying the emotional behavior of a person to classify their emotions, through different sources of information such as facial expressions, voice tone, and brain signals, among others. However, these studies are not available in all situations and have not considered all people and their disabilities. Consequently, this research presents the design of a framework for the recognition of emotions aimed at people with visual disabilities. Firstly, the framework is adapted to this type of people and their disability. Second, short-duration auditory stimuli are used to induce affective states, and a component of acquisition and features extraction of brain signals has been defined. Finally, analysis techniques have been used to model the brain signals, and machine learning models have been evaluated to classify emotions. The expected result is to obtain new evidence on applying a brain-computer interface and affective computing in a scenario not previously evaluated, the recognition of information of the affective states of people with visual disabilities, and the classification of their emotions. The methodology defined to achieve the objective of classifying emotions in people with visual disabilities and the activities for the development of the doctoral thesis are the following: ● State of the art. In this phase, an analysis of related works is carried out on brain-computer interface and affective computing aimed at people with visual disabilities and the classification of their affective states. ● Proposed solution. In this stage, the framework for the classification of emotions in a scenario oriented to people with visual disabilities is presented. Likewise, the elements and components of the framework of each of its phases are described. ● Evaluation and validation. In this section, the activities that allow the evaluation and validation of the framework phases are presented. Additionally, brain signal-based emotion approaches are described to identify which emotions show different behavior in relation to brain signals. ● Documentation and conclusions. In this section, each aspect related to the doctoral thesis is documented and the results that emerge from the research are presented as a conclusion. Based on the results, the framework can classify positive and negative emotions, but the experimentation conducted also shows that the framework's performance depends on the number of features in the dataset and the quality of the brain signals is a determining factor.
Description
Keywords
Emociones, Aprendizaje automático, Computación afectiva, Interacción Hombre Máquina, Discapacidad visual, Personas con discapacidad, Interfaz cerebro-ordenador