|Title:||Semi-supervised learning for speech emotion recognition|
|Advisors:||Mak, M. W. (EIE)|
|Subject:||Hong Kong Polytechnic University -- Dissertations|
Pattern recognition systems
Speech processing systems
|Department:||Faculty of Engineering|
|Pages:||x, 40 pages : illustrations|
|Abstract:||Emotion recognition is an important field with extensive research and applications. Speech emotion recognition is also an irreplaceable part of human-computer interaction (HCI), which has been widely applied in daily life. At present, most of the speech emotion recognition systems use supervised learning algorithms. But they do not achieve good results. In order to improve the recognition performance, this thesis uses semi-supervised learning for speech emotion recognition. In the method, training is used on both labeled and labeled data, the labeled data is used for training an initial emotion classifier for selecting reliable samples from the unlabeled data. The selected unlabeled data are then added to the labeled data to retrain the classifier to improve the accuracy. To demonstrate the performance of this semi-supervised learning strategy, the performance of deep neural networks (DNN) based on semi-supervised learning is compared with the performance of DNNs trained on the labeled data only. Experiments show that the accuracy of DNNs based on semi-supervised is higher than DNNs without using the augmented data. However, this conclusion does not apply to all cases. In particular, the performance of semi-supervised learning is highly unstable, and accuracy cannot be improved if the size of database is too small.|
|Rights:||All rights reserved|
Files in This Item:
|991022270857003411.pdf||For All Users (off-campus access for PolyU Staff & Students only)||600.2 kB||Adobe PDF||View/Open|
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item: