ABSTRACT
For short duration text prompted speaker verification where the amount of enrollment data is limited for each speaker model, it is hard to obtain a robust speaker representation. In these situations of short utterance speaker verification I-vector/GMM approaches work even worse than traditional GMM-MAP modeling method. GMM/HMM framework content matching is one of the state-of-the-art paradigms for short duration text-dependent speaker verification, in which models for individual lexical such as words, syllables, or phonemes are established for the background and speaker to make up mismatch. However, some of the phonemes do not occur in enrollment but happen in the testing recordings, and most of the phonemes have different preceding and succeeding phonemes, both of which leads to coarticulation difference. These are called lexical and context mismatch. In this work, to overcome the data sparceness caused lexical mismatch and context mismatch, phoneme states are clustered applying Earth Mover's Distance and Cauchy-Schwarz divergence as metrics. Performance improved as EER lowered by 6.2%, minDCF08 lowered by 1.9% for Earth Mover's Distance metric, and EER lowered by 3.7%, minDCF08 rised 1.9% for Cauchy-Schwarz divergence metric.
- Najim DehakPatrick J. KennyReda DehakPierre DumouchelPierre Ouellet. 2011. Front-End Factor Analysis for Speaker Verification. IEEE TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING. 19, 4. 788--798. 445 HOES LANE, PISCATAWAY, NJ 08855-4141 USA. Google ScholarDigital Library
- Achintya Kr. SarkarZheng-Hua Tan. 2016. Text Dependent Speaker Verification Using Un-supervised HMM-UBM and Temporal GMM-UBM. 425--429.Google Scholar
- Dong WangLantian LiZhiyuan TangThomas Fang Zheng. 2017. Deep Speaker Verification: Do We Need End to End? 177--181.Google Scholar
- Chao LiXiaokong MaBing JiangXiangang LiXuewei ZhangXiao LiuYing CaoAjay KannanZhenyao Zhu. 2017. Deep Speaker: an End-to-End Neural Speaker Embedding System.Google Scholar
- Jinxi GuoUsha Amrutha NookalaAbeer Alwan. 2017. CNN-Based Joint Mapping of Short and Long Utterance i-Vectors for Speaker Verification Using Short Utterances. 3712--3716.Google Scholar
- Yi LiuLiang HeJia LiuMichael T. Johnson. 2017. Investigation of Frame Alignments for GMM-based Text-prompted Speaker Verification.Google Scholar
- Anthony LarcherAik Lee KongBin MaHaizhou Li. 2014. Text-dependent speaker verification: Classifiers, databases and RSR2015. Speech Communication. 60, 3. 56--77.Google Scholar
- A. LarcherK. A. LeeB. MaH. Li. 2012. The RSR2015: Database for text-dependent speaker verification using multiple pass-phrases. Interspeech.Google Scholar
- Nicolas Lei Yun Scheffer. 2014. Content matching for short duration speaker recognition. INTERSPEECH-2014. 1317--1321.Google Scholar
- Kong Aik LeeAnthony LarcherHelen ThaiBin MaHaizhou Li. 2011. Joint Application of Speech and Speaker Recognition for Automation and Security in Smart Home. 3317--3318.Google Scholar
- Lee, Kong Aik. Larcher, Anthony. Wang, Guangsen. etc (2015): "The reddots data collection for speaker recognition", In INTERSPEECH-2015, 2996--3000.Google Scholar
- S. J. YoungG. EvermannM. J. F. GalesT. HainD. KershawX. LiuG. MooreJ. OdellD. OllasonD. Povey. 2006. The HTK book (v3. 4).Google Scholar
- Robert JenssenDeniz ErdogmusKenneth E. HildJose C. PrincipeTorbjørn Eltoft. 2005. Optimizing the Cauchy-Schwarz PDF Distance for Information Theoretic, Non-parametric Clustering. 34--45. Google ScholarDigital Library
- Peihua LiQilong WangLei Zhang. 2014. A Novel Earth Mover's Distance Methodology for Image Matching with Gaussian Mixture Models. 1689--1696. Google ScholarDigital Library
- Douglas A. ReynoldsThomas F. QuatieriRobert B. Dunn. 2000. Speaker Verification Using Adapted Gaussian Mixture Models.Google Scholar
- Guangsen WangAik Lee KongTrung Hieu NguyenHanwu SunBin Ma. 2016. Joint Speaker and Lexical Modeling for Short-Term Characterization of Speaker. 415--419.Google Scholar
- W. M. CampbellD. E. SturimD. A. ReynoldsA. Solomonoff. 2006. SVM Based Speaker Verification using a GMM Supervector Kernel and NAP Variability Compensation.Google Scholar
- Wei WuThomas Fang ZhengMing Xing XuFrank K. Soong. 2007. A Cohort-Based Speaker Model Synthesis for Mismatched Channels in Speaker Verification. IEEE Transactions on Audio Speech & Language Processing. 15, 6. 1893--1903. Google ScholarDigital Library
- K. KampaE. HasanbelliuJ. C. Principe. 2011. Closed-form cauchy-schwarz PDF divergence for mixture of Gaussians. 2578--2585.Google Scholar
- Fan WangLeonidas J. Guibas. 2012. Supervised Earth Mover's Distance Learning and Its Computer Vision Applications. Springer Berlin Heidelberg.Google Scholar
- Y. RubnerC. TomasiL. J. Guibas. 1998. A Metric for Distributions with Applications to Image Databases. 59. Google ScholarDigital Library
- T. StadelmannB. Freisleben. 2006. Fast and Robust Speaker Clustering Using the Earth Mover'S Distance and Mixmax Models.Google Scholar
Index Terms
- Text Prompted Speaker Verification Based on Phoneme Clustering with Earth Mover's Distane and Cauchy-Schwarz Divergence
Recommendations
Text-Independent/Text-Prompted Speaker Recognition by Combining Speaker-Specific GMM with Speaker Adapted Syllable-Based HMM
We presented a new text-independent/text-prompted speaker recognition method by combining speaker-specific Gaussian Mixture Model (GMM) with syllable-based HMM adapted by MLLR or MAP. The robustness of this speaker recognition method for speaking style'...
Nonparametric Speaker Recognition Method Using Earth Mover's Distance
In this paper, we propose a distributed speaker recognition method using a nonparametric speaker model and Earth Mover's Distance (EMD). In distributed speaker recognition, the quantized feature vectors are sent to a server. The Gaussian mixture model (...
Concatenated phoneme models for text-variable speaker recognition
ICASSP'93: Proceedings of the 1993 IEEE international conference on Acoustics, speech, and signal processing: speech processing - Volume IIThis paper investigates methods that create models to specify both speaker and phonetic information accurately by using only a small amount of training data for each speaker. For a text-dependent speaker recognition method, in which arbitrary key texts ...
Comments