Emotion recognition through facial expressions and speech using machine learning techniques
Loading...
Date
2024-07-10
Authors
Νικολακέας, Ευστράτιος
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The biggest percentage of human communication is nonverbal, it comprises
elements such as facial expressions, speech tone, body position, etc. In human-computer
interaction this critical portion is often missing, resulting in less empathetic and unintuitive
systems. Emotion recognition, a multidisciplinary field, can play a key role in bridging that
gap and developing more robust systems. This thesis constitutes an exploration attempt into
emotion recognition to create a high-performing model, leveraging machine learning
techniques.
The research rests upon two primary pillars, each tackling emotion recognition
through a different aspect. First, the focus will be on facial expressions and the information
it can be extracted from them using deep convolutional neural networks. Continuing,
experimentation with different models for speech emotion recognition will occur alongside
work on improving these models to achieve an adequate level of precision. Having ensured
satisfactory results, the next step is to go through the process of combining these models
into a hybrid, more complete, solution and analyze the challenges that this endeavor
contains.
Summarizing, the findings will be compared to other works on the audiovisual
spectrum of emotion recognition. Future improvements and potential enhancements will be
suggested, along with the significance of the progress achieved this far.
Description
Keywords
Machine learning, Emotion recognition, Convolutional neural networks, Deep learning