Iterative PLDA Adaptation for Speaker Diarization

Lan, Gaël Le; Charlet, Delphine; Larcher, Anthony; Meignier, Sylvain

doi:10.21437/interspeech.2016-572

Cited by 11 publications

(9 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [21], the authors used HAC for speaker linking, but our preliminary experiments with TR scoring using HAC gave inconclusive results. As shown in the lower part of Figure 5, HAC gives erratic variations in terms of DER depending on the number of epochs, for BF M .…”

Section: Diarization Resultsmentioning

confidence: 99%

“…Contrary to the speaker diarization and linking framework of [21] where PLDA is used to compute similarities between ivectors, we decide to replace PLDA by a Neural Network approach. The method is inspired by that of [15] and [16], where the triplet loss [17] is used to train a neural network embedding, which aims at separating faces or speakers.…”

Section: Triplet Ranking Frameworkmentioning

confidence: 99%

See 1 more Smart Citation

A Triplet Ranking-Based Neural Network for Speaker Diarization and Linking

Lan¹,

Charlet²,

Larcher³

et al. 2017

Interspeech 2017

Self Cite

View full text Add to dashboard Cite

This paper investigates a novel neural scoring method, based on conventional i-vectors, to perform speaker diarization and linking of large collections of recordings. Using triplet loss for training, the network projects i-vectors in a space that better separates speakers in terms of cosine similarity. Experiments are run on two French TV collections built from REPERE [1] and ETAPE [2] campaigns corpora, the system being trained on French Radio data. Results indicate that the proposed approach outperforms conventional cosine and Probabilistic Linear Discriminant Analysis scoring methods on both within-and cross-recording diarization tasks, with a Diarization Error Rate reduction of 14% in average.

show abstract

Section: Diarization Resultsmentioning

confidence: 99%

Section: Triplet Ranking Frameworkmentioning

confidence: 99%

A Triplet Ranking-Based Neural Network for Speaker Diarization and Linking

Lan¹,

Charlet²,

Larcher³

et al. 2017

Interspeech 2017

Self Cite

View full text Add to dashboard Cite

show abstract

“…Therefore, we recommend an informative AHC initialization method, similar to our previous paper [51]. After using PLDA to compute the log likelihood ratio between two segment i-vectors [34,35], AHC is applied to perform clustering. Using the AHC results, two prior calculation methods, hard prior and soft prior, are proposed [51].…”

Section: Ahc Initializationmentioning

confidence: 99%

Latent class model with application to speaker diarization

Chen

et al. 2019

J AUDIO SPEECH MUSIC PROC.

View full text Add to dashboard Cite

In this paper, we apply a latent class model (LCM) to the task of speaker diarization. LCM is similar to Patrick Kenny's variational Bayes (VB) method in that it uses soft information and avoids premature hard decisions in its iterations. In contrast to the VB method, which is based on a generative model, LCM provides a framework allowing both generative and discriminative models. The discriminative property is realized through the use of i-vector (Ivec), probabilistic linear discriminative analysis (PLDA), and a support vector machine (SVM) in this work. Systems denoted as LCM-Ivec-PLDA, LCM-Ivec-SVM, and LCM-Ivec-Hybrid are introduced. In addition, three further improvements are applied to enhance its performance. 1) Adding neighbor windows to extract more speaker information for each short segment. 2) Using a hidden Markov model to avoid frequent speaker change points. 3) Using an agglomerative hierarchical cluster to do initialization and present hard and soft priors, in order to overcome the problem of initial sensitivity. Experiments on the National Institute of Standards and Technology Rich Transcription 2009 speaker diarization database, under the condition of a single distant microphone, show that the diarization error rate (DER) of the proposed methods has substantial relative improvements compared with mainstream systems. Compared to the VB method, the relative improvements of LCM-Ivec-PLDA, LCM-Ivec-SVM, and LCM-Ivec-Hybrid systems are 23.5%, 27.1%, and 43.0%, respectively. Experiments on our collected database, CALLHOME97, CALLHOME00 and SRE08 short2-summed trial conditions also show that the proposed LCM-Ivec-Hybrid system has the best overall performance.

show abstract

“…To address it, we recommend a more robust and informative AHC initialization method. After using PLDA to compute the log likelihood ratio between two segment i-vectors [18,19], AHC is applied to get the clustering results. Based on the AHC results, two prior calculation methods, hard prior and soft prior, are proposed.…”

Section: Ahc Initializationmentioning

confidence: 99%

Latent Class Model for Single Channel Speaker Diarization

He¹,

Chen²,

Xu³

et al. 2018

The Speaker and Language Recognition Workshop (Odyssey 2018)

View full text Add to dashboard Cite

Inspired by P. Kenny's variational Bayes (VB) method, we derive a latent class model (LCM) for single channel speaker diarization. Similar to the VB method, the LCM uses soft information and avoids premature hard decisions in its iterations. Different from the VB method, the LCM provides an iterative framework for multi-objective optimization and allows a more flexible way to compute the probability that given a speaker, a segment occurs. Based on this model, we propose a latent class model-i-vector-probabilistic linear discriminant analysis (LCM-Ivec-PLDA) system. Besides, as the divided segments are very short, their neighbors are taken into consideration. To overcome the initial sensitivity problem, we use an agglomerative hierarchical cluster (AHC) to do initialization and present hard and soft priors. Experiments on the NIST RT09 speaker diarization database and our collected database show that the proposed systems are superior to the traditional VB system.

show abstract

Iterative PLDA Adaptation for Speaker Diarization

Cited by 11 publications

References 14 publications

A Triplet Ranking-Based Neural Network for Speaker Diarization and Linking

A Triplet Ranking-Based Neural Network for Speaker Diarization and Linking

Latent class model with application to speaker diarization

Latent Class Model for Single Channel Speaker Diarization

Contact Info

Product

Resources

About