A novel scheme for speaker recognition using a phonetically-aware deep neural network

By Yun Lei, Nicolas Scheffer, Luciana Ferrer, Mitchell McLaren. 2014

627 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

Exploring the Encoding Layer and Loss Function in End-to-End Speaker and Language Recognition System (2018, methodology): “…orderless center components (or units, words, clusters, etc.), 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].”
End-to-end text-dependent speaker verification (2015, background): “Hybrid approaches that include deep learning based components have also proved to be beneficial for text-independent speaker recognition [12, 13, 14].”
Conv-TasNet: Surpassing Ideal Time-Frequency Magnitude Masking for Speech Separation (2018, background): “n devices. Conv-TasNet can also serve as a front-end module for tandem systems in other audio processing tasks, such as multitalker speech recognition [61], [62], [63], [64] and speaker identiﬁcation [65], [66]. On the other hand, several limitations of TasNet must be addressed before it can be actu”
Deep Speaker: an End-to-End Neural Speaker Embedding System (2017, methodology): “This framework can be decomposed into three stages [4]: Step 1 Collect sufﬁcient statistics Step 2 Extract speaker embeddings (i-vector) Step 3 Classify (PLDA) Sufﬁcient statistics (also known as Baum-Welch statistics) are computed from a Gaussian Mixture Model-Universal Background Model (GMM-UBM),…”
Deep Neural Network Approaches to Speaker and Language Recognition (2015, methodology): “The bottleneck layer uses a linear activation and behaves very much like a LDA or PCA transformation on the activation of the previous layer [15], [19].”
Deep Speaker Feature Learning for Text-Independent Speaker Verification (2017, neutral): “For example, the DNN-based i-vector method [6, 7].”
Attentive Statistics Pooling for Deep Speaker Embedding (2018, methodology): “In early studies [2, 3], deep neural networks (DNNs) derived from acoustic models for ASR have been employed as a universal background model (UBM) to provide phoneme posteriors as well as bottleneck features, which are used for, respectively, zeroth- and first-order statistics in i-vector extraction”
Didispeech: A Large Scale Mandarin Speech Corpus (2020, background): “Considering these limitations, we release a novel Mandarin speech database, called DiDiSpeech, which is designed for various speech processing tasks including ASR, TTS, SID, etc. DiDiSpeech consists of two parts: DiDiSpeech-1 and DiDiSpeech-2.”
Self-Attentive Speaker Embeddings for Text-Independent Speaker Verification (2018, background): “Also, hybrid approaches that incorporate deep neural networks (DNNs) trained for automatic speech recognition (ASR) into the i-vector system have proved to be beneficial in some conditions [6, 7, 8, 9, 10].”
Speaker recognition based on deep learning: An overview (2020, methodology): “In addition, the supervised GMM-UBM in (7) can also get 20% relative improvement according to the fourth row.”
Generalized I-vector Representation with Phonetic Tokenizations and Tandem Features for both Text Independent and Text Dependent Speaker Verification (2015, methodology): “Since the features for token extraction and the features for first-order statistics calculation are not necessary the same [14], we show that in terms of first-order statistics calculation, MFCC is superior than tandem features for speaker verification.”
A unified deep neural network for speaker and language recognition (2015, methodology): “This approach has been shown to give signiﬁcant gains for both SR and LR [6, 7, 17].”
Deep bottleneck network based i-vector representation for language identification (2015, methodology): “This validates that the pre-trained DNN for ASR, with a well-annotated corpus, will provide better frame alignment for sufficient statistics calculation.”
End-to-End attention based text-dependent speaker verification (2016, methodology): “More recently, the deep neural networks (DNNs) [12] and recurrent neural networks (RNNs) [13] have been successfully applied to text-dependent speaker verification.”
Deep neural network-based speaker embeddings for end-to-end speaker verification (2016, positive): “Recently, this paradigm has been improved by incorporating deep neural network (DNN) acoustic models [8, 9, 10, 11, 12].”
Improving speaker recognition performance in the domain adaptation challenge using deep neural networks (2014, methodology): “The posteriors of the DNN-based system are used to compute the SS and to define an ancillary UBM needed for the i-vector computation [20, 22].”
Speaker verification and spoken language identification using a generalized i-vector framework with phonetic tokenizations and tandem features (2014, methodology): “Since the features for extracting tokens and the features for calculating the ﬁrst-order statistics are not necessary the same [10], we show that in terms of ﬁrst-order statistics calculation, MFCC is superior than tandem features for SV, and vice versa 10 .”