Multisource I-Vectors Domain Adaptation Using Maximum Mean Discrepancy Based Autoencoders
- 24 August 2018
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ACM Transactions on Audio, Speech, and Language Processing
- Vol. 26 (12) , 2412-2422
Abstract
Like many machine learning tasks, the performance of speaker verification (SV) systems degrades when training and test data come from very different distributions. What's more, both training and test data themselves could be composed of heterogeneous subsets. These multisource mismatches are detrimental to SV performance. This paper proposes incorporating maximum mean discrepancy (MMD) into the loss function of autoencoders to reduce these mismatches. MMD is a nonparametric method for measuring the distance between two probability distributions. With a properly chosen kernel, MMD can match up to infinite moments of data distributions. We generalize MMD to measure the discrepancies of multiple distributions. We call the generalized MMD domainwise MMD. Using domainwise MMD as an objective function, we propose two autoencoders, namely nuisance-attribute autoencoder (NAE) and domain-invariant autoencoder (DAE), for multisource i-vector adaptation. NAE encodes the features that cause most of the multisource mismatch measured by domainwise MMD. DAE directly encodes the features that minimize the multisource mismatch. Using these MMD-based autoencoders as a preprocessing step for PLDA training, we achieve a relative improvement of 19.2% EER on the NIST 2016 SRE compared to PLDA without adaptation. We also found that MMD-based autoencoders are more robust to unseen domains. In the domain robustness experiments, MMD-based autoencoders show 6.8% and 5.2% improvements over IDVC on female and male Cantonese speakers, respectively.Keywords
Funding Information
- Research Grants Council of Hong Kong SAR (PolyU 152137/17E, PolyU 152518/16E)
- Ministry of Science and Technology, Taiwan (107-2634-F-009-003)
This publication has 29 references indexed in Scilit:
- Deep semi-supervised learning for domain adaptation2015
- FaceNet: A unified embedding for face recognition and clustering2015
- Inter dataset variability compensation for speaker recognition2014
- Unsupervised adaptation of PLDA by using variational Bayes methods2014
- A study of voice activity detection techniques for NIST speaker recognition evaluationsComputer Speech & Language, 2013
- Learning Deep Architectures for AIFoundations and Trends in Machine Learning, 2009
- Reducing the Dimensionality of Data with Neural NetworksScience, 2006
- Learning a Similarity Metric Discriminatively, with Application to Face Verification2005
- On the limited memory BFGS method for large scale optimizationMathematical Programming, 1989
- Auto-association by multilayer perceptrons and singular value decompositionBiological Cybernetics, 1988