University of the Aegean
Νέες τεχνικές τεχνητής νοημοσύνης και μετασχηματισμού δεδομένων για την αύξηση απόδοσης αλγορίθμων ταξινόμησης
Abstract
dc:descriptionThe present doctoral thesis effort contributes in two different scientific fields. The first isthe Emotion Recognition field and the second is a field of Machine Learning, the DataPreprocessing-transformation methods in classification problems in order to boost theclassification results.In this research we deal with three major issues of Emotion Recognition fromspeech field: the available emotion databases and their usability in emotion recognitionfrom speech, the various procodic features selection methods and evaluating theircontribution to emotion recognition from speech and we present classification rates ofvarious classifiers that have been used in emotion recognition field, according to theliterature.In the scientific field of Machine Learning, the main contribution of the presentthesis is that systematic research was conducted to develop and evaluate an innovativepre-processing-transformation data method. The proposed method was tested in fourwell-known and difficult classification problems. This technique transforms the originaldata in such a way that minimize the redundancy of the data and reducing the number ofinput variables.The conclusion of this research is that the proposed method significantlyimproves the classification results and reduces the classification time in all the data setsthat was tested.
Degree
thesis:*- Grantor dc:publisher
- University of the Aegean
- Year dc:date
- 2016
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Iliou, Theodoros
- Ηλίου, Θεόδωρος
Subjects
dc:subject × 6Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/37448