Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor | Department of Electronic and Information Engineering | en_US |
| dc.contributor.advisor | Mak, M. W. (EIE) | - |
| dc.creator | Pang, Xiao-min | - |
| dc.identifier.uri | https://theses.lib.polyu.edu.hk/handle/200/8146 | - |
| dc.language | English | en_US |
| dc.publisher | Hong Kong Polytechnic University | - |
| dc.rights | All rights reserved | en_US |
| dc.title | Fusion of SNR-dependent PLDA models for noise robust speaker verification | en_US |
| dcterms.abstract | Speaker verification is to verify the identify of a claimant based on his or her voice. In many speaker verification systems, the i-vector representation and PLDA classifier have shown state-of-the-art performance. However, in real-world environments, additive and convolutive noise cause loss of speaker-specific information and mismatches between training and recognition conditions, degrading the performance. The objective of this thesis is to enhance the robustness of speaker verification systems by combining multi-condition PLDA and mixture of SNR-dependent PLDA. Both hard-and soft-decision strategies are employed for the SNR-dependent PLDA. With hard-decisions, the SNR of test utterances is directly used to determine the best SNR-dependent PLDA model to score against the target-speaker's i-vectors. With soft-decisions, the posterior probabilities of the SNR of a test utterance determine the weights to be applied to the SNR-dependent PLDA scores for computing the final verification score. Linear fusion and logistic regression fusion are used for the fusion system. With linear fusion, a multi-condition PLDA model and a mixture of SNR-dependent PLDA models are combined by a predefined fusion weight. With logistic regression fusion, multiple systems can be combined by the fusion parameters that are derived from development data. The performance of the fusion system and the SNR-dependent system is evaluated on the NIST 2012 SRE. Results show that the SNR-dependent PLDA models can reduce EER and that the fusion system is more robust than the conventional i-vector/PLDA systems under noisy conditions. It is also found that the fusion systems are insensitive to decision strategies, decision thresholds and Z-norm parameters. | en_US |
| dcterms.extent | x, 52 leaves : illustrations (some color) ; 30 cm | en_US |
| dcterms.isPartOf | PolyU Electronic Theses | en_US |
| dcterms.issued | 2014 | en_US |
| dcterms.educationalLevel | All Master | en_US |
| dcterms.educationalLevel | M.Sc. | en_US |
| dcterms.LCSH | Automatic speech recognition. | en_US |
| dcterms.LCSH | Hong Kong Polytechnic University -- Dissertations | en_US |
| dcterms.accessRights | restricted access | en_US |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| b28194871.pdf | For All Users (off-campus access for PolyU Staff & Students only) | 11.82 MB | Adobe PDF | View/Open |
Copyright Undertaking
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:
https://theses.lib.polyu.edu.hk/handle/200/8146