Date of Award
3-2018
Document Type
Dissertation
Degree Name
Doctor of Philosophy
Department
Computer Sciences
First Advisor
Prof. Paul Walsh
Second Advisor
Dr. Kieran Delaney
Third Advisor
Prof. Matthias Hemmje
Abstract
The research explores the opportunities, challenges, limitations, and presents advancements in computing that relates to, arises from, or deliberately influences emotions (Picard, 1997). The field is referred to as Affective Computing (AC) and is expected to play a major role in the engineering and development of computationally and cognitively intelligent systems, processors and applications in the future. Today the field of AC is bolstered by the emergence of multiple sources of affective data and is fuelled on by developments under various Internet of Things (IoTs) projects and the fusion potential of multiple sensory affective data streams. The core focus of this thesis involves investigation into whether the sensitivity and specificity (predictive performance) of AC, based on the fusion of multi-sensor data streams, is fit for purpose? Can such AC powered technologies and techniques truly deliver increasingly accurate emotion predictions of subjects in the real world? The thesis begins by presenting a number of research justifications and AC research questions that are used to formulate the original thesis hypothesis and thesis objectives. As part of the research conducted, a detailed state of the art investigations explored many aspects of AC from both a scientific and technological perspective. The complexity of AC as a multi-sensor, multi-modality, data fusion problem unfolded during the state of the art research and this ultimately led to novel thinking and origination in the form of the creation of an AC conceptualised architecture that will act as a practical and theoretical foundation for the engineering of future AC platforms and solutions. The AC conceptual architecture developed as a result of this research, was applied to the engineering of a series of software artifacts that were combined to create a prototypical AC multi-sensor platform known as the Emotion Fusion Server (EFS) to be used in the thesis hypothesis AC experimentation phases of the research. The thesis research used the EFS platform to conduct a detailed series of AC experiments to investigate if the fusion of multiple sensory sources of affective data from sensory devices can significantly increase the accuracy of emotion prediction by computationally intelligent means. The research involved conducting numerous controlled experiments along with the statistical analysis of the performance of sensors for the purposes of AC, the findings of which serve to assess the feasibility of AC in various domains and points to future directions for the AC field. The AC experiments data investigations conducted in relation to the thesis hypothesis used applied statistical methods and techniques, and the results, analytics and evaluations are presented throughout the two thesis research volumes. The thesis concludes by providing a detailed set of formal findings, conclusions and decisions in relation to the overarching research hypothesis on the sensitivity and specificity of the fusion of vision and wearables sensor modalities and offers foresights and guidance into the many problems, challenges and projections for the AC field into the future.
Recommended Citation
Keary, Alphonsus, "Affective Computing for Emotion Detection using Vision and Wearable Sensors" (2018). PhDs [online].
Available at: https://sword.cit.ie/scidiss/3
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Volume 2 of 2 Thesis
Access Level
info:eu-repo/semantics/openAccess