Speaker Recognition The A TVS-UAM System at NIST SRE 05

González-Rodríguez, Joaquín; Ramos-Castro, Daniel; Toledano, Doroteo T.; Montero-Asenjo, Alberto; Gónzalez-Domínguez, Javier; López-Moreno, Ignacio; Fiérrez-Aguilar, J.; Garcia‐Romero, Daniel; Ortega‐Garcia, Javier

doi:10.1109/maes.2007.327534

Cited by 6 publications

(4 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Actually the system is the sum fusion of two SVM systems, one using 19 MFCC coefficients plus deltas and the other using SDC-MFCCs (7-2-3-7) [14]. In order to avoid channel mismatch effects, Cepstral Mean Normalization is applied, followed by RASTA filtering and feature mapping [15]. Both systems use a kernel expansion on the whole observation sequence, and a separating hyperplane is computed between the target language features and the background model.…”

Section: Svm Sytems With Mfcc and Sdc-mfcc Featuresmentioning

confidence: 99%

Improved language recognition using better phonetic decoders and fusion with MFCC and SDC features

Toledano¹,

Gónzalez-Domínguez²,

Abejón-Gonzalez³

et al. 2007

Interspeech 2007

Self Cite

View full text Add to dashboard Cite

One of the most popular and better performing approaches to language recognition (LR) is Parallel Phonetic Recognition followed by Language Modeling (PPRLM). In this paper we report several improvements in our PPRLM system that allowed us to move from an Equal Error Rate (EER) of over 15% to less than 8% on NIST LR Evaluation 2005 data still using a standard PPRLM system. The most successful improvement was the retraining of the phonetic decoders on larger and more appropriate corpora. We have also developed a new system based on Support Vector Machines (SVMs) that uses as features both Mel Frequency Cepstral Coefficients (MFCCs) and Shifted Delta Cepstra (SDC). This new SVM system alone gives an EER of 10.5% on NIST LRE 2005 data. Fusing our PPRLM system and the new SVM system we achieve an EER of 5.43% on NIST LRE 2005 data, a relative reduction of almost 66% from our baseline system.

show abstract

Section: Svm Sytems With Mfcc and Sdc-mfcc Featuresmentioning

confidence: 99%

Improved language recognition using better phonetic decoders and fusion with MFCC and SDC features

Toledano¹,

Gónzalez-Domínguez²,

Abejón-Gonzalez³

et al. 2007

Interspeech 2007

Self Cite

View full text Add to dashboard Cite

show abstract

“…Feature extraction obtains 19 MFCC coefficients plus deltas. In order to avoid channel mismatch effects, cepstral mean normalization is applied, followed by RASTA filtering and feature mapping (see [4] for details). The similarity computation is based on SVC [3].…”

Section: Baseline Systemmentioning

confidence: 99%

“…Speaker verification has been dominated in the last decade by systems working at the spectral level of the speaker identity [1]. Techniques like Gaussian Mixture Models (GMM) [2] or Support Vector Machines (SVM) using Generalized Linear Discriminant Sequence (GLDS) kernels [3] have demonstrated its superiority to higher level approaches [1,4]. In recent years, hybrid approaches such as GMM-SVM systems [5] and channel compensation techniques like factor analysis [6] or nuisance attribute projection [7] have led to a significant improvement of the state-of-the-art performance.…”

Section: Introductionmentioning

confidence: 99%

Support vector regression for speaker verification

López-Moreno¹,

Mateos-Garcia²,

Ramos³

et al. 2007

Interspeech 2007

Self Cite

View full text Add to dashboard Cite

This paper explores Support Vector Regression (SVR) as an alternative to the widely-used Support Vector Classification (SVC) in GLDS (Generalized Linear Discriminative Sequence)based speaker verification. SVR allows the use of a ε-insensitive loss function which presents many advantages. First, the optimization of the ε parameter adapts the system to the variability of the features extracted from the speech. Second, the approach is robust to outliers when training the speaker models. Finally, SVR training is related to the optimization of the probability of the speaker model given the data. Results are presented using the NIST SRE 2006 protocol, showing that SVR-GLDS yields a relative improvement of 31% in EER compared to SVC-GLDS.

show abstract

“…Some spectral characteristics in the speech signal such as formant distribution and variation are related to speaker-dependent characteristics. Systems exploiting acoustic information are based on the short-term spectral identity information and this spectral information may be analyzed in order to recognize the speaker identity [1]. In present rapid development and during putting automated systems into existence they may be applied also in the devices for speaker recognition, for instance authorised check in, banking via telephone, and in the telephone credit cards [2].The goal of speaker recognition is to automatically extract information from speech signal.…”

Section: Introductionmentioning

confidence: 99%

Comparison of wavelet packets and DWT in a gender based multichannel speaker recognition system

Jayakumar

Anto

2013

2013 Ieee Conference on Information and Communication Technologies

View full text Add to dashboard Cite

As the technology develops, people looks forward more on Speech analysis for Speaker Recognition. A large multichannel corpus, including mobile phone, mobile phone headset, laptop microphone and laptop headset, is collected to evaluate the system performance. Also, a comparative study among wavelet packets and DWT on a gender based Speaker recognition database is evaluated in this paper. This work is carried out with a database of four hundred and eighty samples. Speech samples were collected through multi channels devices such as laptop microphone and mobile head set. The speech data used in this work is commonly used for banking purpose such as name of the person, Account Number, Password and two security questions. A neural network Classifier is used for classification.

show abstract

Speaker Recognition The A TVS-UAM System at NIST SRE 05

Cited by 6 publications

References 13 publications

Improved language recognition using better phonetic decoders and fusion with MFCC and SDC features

Improved language recognition using better phonetic decoders and fusion with MFCC and SDC features

Support vector regression for speaker verification

Comparison of wavelet packets and DWT in a gender based multichannel speaker recognition system

Contact Info

Product

Resources

About