Speaker adaptation of context dependent deep neural networks

By Hank Liao. 2013

245 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

In Signal Processing

Citations in the corpus

Multiple-Source Adaptation with Domain Classifiers (2020, background): “In speech recognition, each domain may correspond to a different group of speakers and an acoustic model learned for each domain may be available, and the problem consists of devising a general recognizer for a broader population, a mixture of the source domains [Liao, 2013].”
Algorithms and Theory for Multiple-Source Adaptation (2018, background): “Such problems arise commonly in speech recognition where different groups of speakers (domains) yield different acoustic models and the problem is to derive an accurate acoustic model for a broader population that may be viewed as a mixture of the source groups [Liao, 2013].”
Deep Learning: Methods and Applications (2014, methodology): “Several studies were carried out on unsupervised or semi-supervised adaptation of DNN acoustic models with different types of input features with success [223, 405].”
EESEN: End-to-end speech recognition using deep RNN models and WFST-based decoding (2015, methodology): “We will study new speaker adaptation [35, 36] and adaptive training [37, 38] techniques for the CTC models.”
The CAPIO 2017 Conversational Speech Recognition System (2017, methodology): “As a result there have been alternative approaches to update only a small part of a neural network model [7, 8, 9, 10] to obtain adaptation benefits.”
Explicit Inductive Bias for Transfer Learning with Convolutional Networks (2018, methodology): “All the experiments are performed with Tensorﬂow (Abadi et al. 2015).”
Multiple-source adaptation theory and algorithms (2020, background): “Such problems arise commonly in speech recognition where different groups of speakers (domains) yield different acoustic predictors and the problem is to derive an accurate acoustic predictor for a broader population that may be viewed as a mixture of the source groups [37].”
Large-Scale Domain Adaptation via Teacher-Student Learning (2017, unknown): “Because it is typically very time-consuming or expensive to transcribe large amounts of data for a new domain, domain-adaptation approaches have been proposed to bootstrap the training of a new system from an existing welltrained model [5][6][7][8].”
On speaker adaptation of long short-term memory recurrent neural networks (2015, unknown): “First, the speaker-independent (SI) model, or certain layers of the model, are re-updated on the adaptation data of each testing speaker [18, 19].”
Unsupervised Adaptation of Acoustic Models for ASR Using Utterance-Level Embeddings from Squeeze and Excitation Networks (2019, background): “There are also approaches that directly fine-tune the entire speaker-independent (SI) DNN acoustic model or certain layers to make it speaker adapted [7], [8].”
The Automatic Speech recogition In Reverberant Environments (ASpIRE) challenge (2015, methodology): “Evaluation data was used for semisupervised retraining [29].”
Embedding-Based Speaker Adaptive Training of Deep Neural Networks (2017, methodology): “Or, alternatively, the input or output layer is adapted by an affine transformation estimated using the adaptation data [6][7][8].”
Transfer learning for speech and language processing (2015, methodology): “Some comparison for adaptation on different components can be found in [98], [99].”
Factorized Hidden Layer Adaptation for Deep Neural Network Based Acoustic Modeling (2016, methodology): “Moreover, some approaches perform the adaptation on a subset of parameters, including the last hidden layer [45], [46], input layer [46], biases [47], output layer biases [48], or more active hidden units of the network [45].”
Unsupervised adaptation with domain separation networks for robust speech recognition (2017, methodology): “To address this issue, regularization-based approaches are proposed in [9, 10, 11, 12] to regularize the neuron output distributions or the model parameters.”
Recent progresses in deep learning based acoustic models (2017, methodology): “Although unsupervised adaptation methods can be used by generating labels from a decoder, the performance gap be-tween supervised and unsupervised adaptation is large [98].”
Learning hidden unit contributions for unsupervised speaker adaptation of neural network acoustic models (2014, methodology): “Explicit adaptation to speaker or acoustic characteristics can further improve accuracy [12, 13, 14, 15, 16].”
Fast Adaptation of Deep Neural Network Based on Discriminant Codes for Speech Recognition (2014, background): “In [19], it has explored how to adapt DNN to new speakers by other retraining and regularization tricks.”
The NTT CHiME-3 system: Advances in speech enhancement and recognition for mobile multi-microphone devices (2015, methodology): “Adaptation is performed by re-training the neural network with SGD on a modest number (i.e., 1800) of mini-batches with a ﬁ xed learning rate of 0.02 [36].”