Joint Factor Analysis Versus Eigenchannels in Speaker Recognition

By Patrick Kenny, Gilles Boulianne, Pierre Ouellet, Pierre Dumouchel. 2007

789 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Jinyu Li
2. Shinji Watanabe
3. Dong Yu
4. Yoshua Bengio
5. Li Deng
6. Yifan Gong
7. Zhuo Chen
8. Haizhou Li

In Signal Processing

1. Jinyu Li
2. Shinji Watanabe
3. Patrick Kenny
4. Yifan Gong
5. Dong Yu
6. Najim Dehak
7. Zhuo Chen
8. Haizhou Li

Citations in the corpus

• End-to-end text-dependent speaker verification (2015, background): “State-of-the art representations are a summary of frame-level information, such as i-vectors [7, 8] and d-vectors (Section 3).”
• Deep Speaker: an End-to-End Neural Speaker Embedding System (2017, background): “Though many algorithms have pushed the state-of-the-art over the past couple years [1][2][3][4][5][6][7], speaker recognition is still a challenging task.”
• CN-Celeb: A Challenging Chinese Speaker Recognition Dataset (2019, background): “JFA [3, 4] and PLDA [5] are the most famous among such models.”
• Deep neural networks for small footprint text-dependent speaker verification (2014, methodology): “Background models vary from simple Gaussian mixture model (GMM) based Universal Background Models (UBMs) [1] to more sophisticated Joint Factor Analysis (JFA) based models [2, 3, 4].”
• Teacher-Student Training for Text-Independent Speaker Recognition (2018, background): “Joint factor analysis [3] is an example of generative approaches that explicitly models speaker– and channel– dependent factors.”
• Non-intrusive Speech Quality Assessment Using Neural Networks (2019, methodology): “As in the Joint Factor Analysis (JFA) [20], the i-vector framework also considers that speaker and channel variability lies in a lower subspace of the GMM supervectors [21].”
• Speaker recognition based on deep learning: An overview (2020, methodology): “The i-vector-PLDA technique and its variants have also been successfully used in text-dependent speaker recognition tasks [8, 9, 10].”
• A Review of Speaker Diarization: Recent Advances with Deep Learning (2021, background): “(15) In the JFA approach, the given GMM supervector is decomposed into speaker independent, speaker dependent, channel dependent, and residual components.”
• The IBM 2016 Speaker Recognition System (2016, methodology): “The research trend in this field has gradually evolved from joint factor analysis (JFA) based methods, which attempt to model the speaker and channel subspaces separately [1], towards the i-vector approach that models both speaker and channel variabilities in a single low-dimensional (e.”
• Joint Factor Analysis of Speaker and Session Variability: Theory and Algorithms (2006, methodology): “We also need a speaker-dependent estimation algorithm for the hyperparameters both for constructing likelihood ratio statistics for speaker verification [8], [1] and for progressive speaker adaptation.”
• A Study of Interspeaker Variability in Speaker Verification (2008, methodology): “To estimatev and d we pooled all of the recordings of each speaker in the factor analysis training set and ignored channel effects as in [24].”
• End-to-End attention based text-dependent speaker verification (2016, positive): “Previously, the traditional techniques [7, 8] used for textindependent speaker verification have been found ineffective for the text-dependent tasks [4].”
• Deep bottleneck features for i-vector based text-independent speaker verification (2015, methodology): “The technique is inspired by joint factor analysis (JFA) framework [5, 12], where the Gaussian mixture model (GMM) mean is factorized into the speaker and channel factors using separate subspaces.”
• Deep Speaker Feature Learning for Text-Independent Speaker Verification (2017, methodology): “For example, the famous Gaussian mixture modeluniversal background model (GMM-UBM) framework [1] and the subsequent subspace models, including the joint factor analysis approach [2] and the i-vector model [3].”
• Front-End Factor Analysis for Speaker Verification (2011, methodology): “O VER recent years, Joint Factor Analysis (JFA) [1]–[3] has demonstrated state-of-the-art performance for text-independent speaker detection tasks in the NIST speaker recognition evaluations (SREs).”
• Spoken Language Recognition: From Fundamentals to Practice (2013, methodology): “In acoustic approaches, the latent factor modeling technique, as used in JFA [57], [140] and i-vector [34], is unleashing its potential in language recognition [35].”