Speaker Verification Using Adapted Gaussian Mixture Models

By Douglas A. Reynolds, Thomas F. Quatieri, Robert B. Dunn. 2000

5,008 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Speech & Audio, 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

In Speech & Audio

In Signal Processing

Citations in the corpus

Local Aggregation for Unsupervised Learning of Visual Embeddings (2019, methodology): “Given the need for cluster-based identification, a variety of alternative approaches to k-means are theoretically possible, including DBSCAN [17], Affinity Propagation [19], spectral methods [63, 66], and gaussian mixtures [60, 61].”
Exploring the Encoding Layer and Loss Function in End-to-End Speaker and Language Recognition System (2018, methodology): “), includes vector quantization (VQ) codebooks learned by K-means [12], a universal background model (UBM) learned by Gaussian Mixture Model (GMM) GMM [13, 14] or a supervised phonetically-aware acoustic model learned by DNN [10, 15].”
VoxCeleb: A Large-Scale Speaker Identification Dataset (2017, background): “For a long time, speaker identification was the domain of Gaussian Mixture Models (GMMs) trained on low dimensional feature vectors [7, 8].”
Deep Voice 2: Multi-Speaker Neural Text-to-Speech (2017, methodology): “Speaker classiﬁcation has been studied with both traditional GMM-based methods (e.g., Reynolds et al., 2000) and more recently with deep learning approaches (e.g., Li et al., 2017).”
End-to-end text-dependent speaker verification (2015, methodology): “For example, GMM-UBM [7] or i-vector models does not directly optimize a verification problem; the PLDA [5] model is not followed a refinement of the i-vector extraction; or long contextual features may be ignored by frame-based GMM-UBM models [7].”
Zero-Shot Multi-Speaker Text-To-Speech with State-Of-The-Art Neural Speaker Embeddings (2019, background): “The aggregation of xt is similar to the supervector notion in [29].”
Domain Adaptation for Visual Applications: A Comprehensive Survey (2017, background): “classification [9, 10, 11], sentiment analysis [12, 13], WiFi-based localization [14], speech recognition across different speakers [15, 16], object recognition in images acquired in different conditions [17, 18, 19, 20, 21], video concept detection [22], video event recognition [23], activity recog”
Spoofing and countermeasures for speaker verification: A survey (2015, methodology): “GMMs have been used intensively and their combination with a universal background model (UBM) has become the de facto standard, commonly referred to as the GMM-UBM approach (Reynolds et al., 2000).”
Survey on speech emotion recognition: Features, classification schemes, and databases (2011, methodology): “While HMM is the most widely used classifier in the task of automatic speech recognition (ASR), GMM is considered the state-of-the-art classifier for speaker identification and verification [106].”
Power efficient processor architecture and the cell processor (2005, background): “…have over the last couple of decades been driven by two primary metrics that determine performance: performance per cycle (often approximated by the number of instructions completed per processor cycle), and design frequency (e.g. design cycle time measured in fanout-4 inverter delays).”
Person Authentication Using Brainwaves (EEG) and Maximum A Posteriori Model Adaptation (2007, methodology): “We propose to adopt a statistical framework widely used in other biometric authentication approaches such as speaker authentication [13] or face verification [2].”
The MIT Mobile Device Speaker Verification Corpus: Data Collection and Preliminary Experiments (2006, methodology): “For scenarios involving limited enrollment data, this approach allows the system to smooth phone dependent models that may not be adequately trained with more robustly trained global Gaussian mixture models (GMMs) [10].”
CONSISTENCY OF A RECURSIVE ESTIMATE OF MIXING DISTRIBUTIONS (2009, background): “In recent years, these models have been widely used in genetics, bioinformatics, proteomics, computer vision, speech analysis and a host of other research areas; see, for example, [1, 5, 7, 21, 25, 27, 31].”
Evolving spiking neural networks for audiovisual information processing (2010, methodology): “Based on these properties, in recent years, Vector Quantization (VQ) (Burileanu, Moraru, Bojan, Puchiu, & Stan, 2002; Gray, 1984) and Gaussian Mixture Models (GMM) (Bimbot et al., 2004; Reynolds et al., 2000) became the standard approaches to tackle the text-independent speaker authentication proble”
A study of Gaussian mixture models of color and texture features for image classification and segmentation (2006, background): “Because of these advantages, there are applications, for example speaker verification, in which a diagonal covariance is used [41].”
Support vector machines for speaker and language recognition (2006, methodology): “MAP estimation) of the UBM parameters using the designated training data [14].”
A Tutorial on Text-Independent Speaker Verification (2004, background): “” State-of-the-art speaker recognition systems, widely described in this contribution, provide a fully automated approach, handling huge quantities of speech information at a low-level acoustic signal processing [78, 79, 80].”
A probabilistic approach for damage identification and crack mode classification in reinforced concrete structures (2013, methodology): “GMM is an unsupervised distribution–based classification technique (Bilmes, 1998) that has been successfully used in many fields, including sound recognition (Reynolds et al., 2000; Reynolds and Rose, 1995), image processing (Permuter et al.”
Joint Factor Analysis of Speaker and Session Variability: Theory and Algorithms (2006, methodology): “In the case where u = 0 and v = 0, so that the factor analysis model reduces to the prior for classical MAP, the 'relevance factors' of [6] are the diagonal entries of d −2 Σ (as we will explain).”