An Investigation into Back-end Advancements for Speaker Recognition in Multi-Session and Noisy Enrollment Scenarios

Liu, Gang; Hansen, John H. L.

doi:10.1109/taslp.2014.2352154

Cited by 39 publications

(19 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…I-vector has proved to be an effective representation for the inherent information of speakers (see e.g., [15,16,17]). As previously noted, however, the performance degradation in ASV systems is attributed to the two issues.…”

Section: I-vector In Asvmentioning

confidence: 99%

Centroid-aware local discriminative metric learning in speaker verification

Sheng

Dong

et al. 2017

Pattern Recognition

View full text Add to dashboard Cite

We propose a new mechanism to pave the way for efficient learning against class-imbalance and improve representation of identity vector (i-vector) in automatic speaker verification (ASV). The insight is to effectively exploit the inherent structure within ASV corpus -centroid priori. In particular: 1) to ensure learning efficiency against class-imbalance, the centroid-aware balanced boosting sampling is proposed to collect balanced mini-batch; 2) to strengthen local discriminative modeling on the mini-batches, neighborhood component analysis (NCA) and magnet loss (MNL) are adopted in ASV-specific modifications. The integration creates adaptive NCA (AdaNCA) and linear MNL (LMNL). Numerical results show that LMNL is a competitive candidate for lowdimensional projection on i-vector (EER=3.84% on SRE2008, EER=1.81% on SRE2010), enjoying competitive edge over linear discriminant analysis (LDA). AdaNCA (EER=4.03% on SRE2008, EER=2.05% on SRE2010) also performs well. Furthermore, to facilitate the future study on boosting sampling, connections between boosting sampling, hinge loss and data augmentation have been established, which help understand the behavior of boosting sampling further.

show abstract

Section: I-vector In Asvmentioning

confidence: 99%

Centroid-aware local discriminative metric learning in speaker verification

Sheng

Dong

et al. 2017

Pattern Recognition

View full text Add to dashboard Cite

show abstract

“…Recent SV benchmark datasets, such as the NIST SRE 2012 ( [5]), RSR2015 ( [6]) and Reddots ( [7]), have multiple audio files to enroll target speakers. In order to exploit multiple examples, various methods have been explored [8,9]. As the PLDA scoring framework is a well established back-end classifier, a majority of these methods target to exploit it in order to leverage its effectiveness.…”

mentioning

confidence: 99%

“…Generally, such methods extract an i-vector for each available session. These models are either compared to the test recording either individually or after averaging them to obtain one single model for the target speaker [8,10]. In the former case, the scores need to be combined (for instance, through simple averaging) [11].…”

mentioning

confidence: 99%

Intra-class covariance adaptation in PLDA back-ends for speaker verification

Madikeri

Ferras

Motlíček

et al. 2017

2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

Multi-session training conditions are becoming increasingly common in recent benchmark datasets for both textindependent and text-dependent speaker verification. In the state-of-the-art i-vector framework for speaker verification, such conditions are addressed by simple techniques such as averaging the individual i-vectors, averaging scores, or modifying the Probabilistic Linear Discriminant Analysis (PLDA) scoring hypothesis for multi-session enrollment. The aforementioned techniques fail to exploit the speaker variabilities observed in the enrollment data for target speakers. In this paper, we propose to exploit the multi-session training data by estimating a speaker-dependent covariance matrix and updating the intra-speaker covariance during PLDA scoring for each target speaker. The proposed method is further extended by combining covariance adaptation and score averaging. In this method, the individual examples of the target speaker are compared against the test data as opposed to an averaged ivector, and the scores obtained are then averaged. The proposed methods are evaluated on the NIST SRE 2012 dataset. Relative improvements of up to 29% in equal error rate are obtained.

show abstract

“…As principais contribuições resultantes desta tese são descritas na seção seguinte. [23], sendo atualmente uma técnica amplamente conhecida e explorada por diversos pesquisadores, como em [8], [9], [24], [25], [26]. Estes sistemas vêm sendo utilizados em aplicações biométricas comerciais nas últimas décadas, desde aplicações mais simples como o desbloqueio do smartphone com o sinal de voz ou até mesmo sistemas de segurança nacional como identificação de vozes em perícias criminais [27].…”

Section: Figura 1 -Sistema Biométrico Genéricounclassified

“…O termo é uma abreviação de identity-vector ou vetor-identidade, dado o propósito da aplicação e, desde então, tem sido utilizada como o estado da arte para reconhecimento de locutor [26], [75], [76], [77]. Sua utilização também é baseada na criação prévia de um modelo GMM através da adaptação MAP a partir de um modelo UBM, como descrito na seção 2.3.2.…”

Section: Análise Fatorialunclassified

Biometria multimodal baseada nos sinais de voz e facial

Parada¹

View full text Add to dashboard Cite

An Investigation into Back-end Advancements for Speaker Recognition in Multi-Session and Noisy Enrollment Scenarios

Cited by 39 publications

References 52 publications

Centroid-aware local discriminative metric learning in speaker verification

Centroid-aware local discriminative metric learning in speaker verification

Intra-class covariance adaptation in PLDA back-ends for speaker verification

Biometria multimodal baseada nos sinais de voz e facial

Contact Info

Product

Resources

About