Recognizing emotion in speech using a multilayer perceptron
Keywords
No Thumbnail Available
Authors
Issue Date
2020-01-31
Language
en
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Recognizing human emotion from speech has become an active theme in the field of Human-
Computer Interaction. The demand for Speech Emotion Recognition systems is growing
given its numerous applications. This thesis presents and analyses an implementation of a
model which has the ability to recognize emotions (neutral, calm, happy, sad, angry, fear,
surprise and disgust) from human speech. The approach contains a multilayer perceptron
classifier. It was found that the designed model managed to reach a mean accuracy of 59%
when classifying eight emotions, an accuracy of 75% when classifying four emotions and an
accuracy of 77% when classifying the data according to its emotional valance. This
performance compares favorably with more complex classifiers reported in the literature.
Also, an experiment was conducted where participants classified recordings from the same
dataset. From the obtained results is concluded that the proposed model performs comparable
to humans, which scored a mean performance of 60% on the classifying task, but the model is
more consistent.
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen