Low Latency Acoustic Modeling Using Temporal Convolution and LSTMs

By Vijayaditya Peddinti, Yiming Wang, Daniel Povey, Sanjeev Khudanpur. 2018

171 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Model Architectures, 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 Model Architectures

In Speech & Audio

In Signal Processing

Citations in the corpus

TED-LIUM 3: twice as much data and corpus repartition for experiments on speaker adaptation (2018, methodology): “In section 5, we report experimental results obtained on the speaker adaptation distribution by exploiting GMM-HMM and TDNN-Long Short-Term Memory (TDNN-LSTM) [6] acoustic models and two standard adaptation techniques (i-vectors and feature space maximum linear regression (fMLLR)).”
Snips Voice Platform: an embedded Spoken Language Understanding system for private-by-design voice interfaces (2018, methodology): “We train a deep neural network, consisting of time-delay layers similar to those presented in [39], and long short-term memory layers similar to those of [40].”
Semi-Orthogonal Low-Rank Matrix Factorization for Deep Neural Networks (2018, methodology): “2 we will show a comparison between our previous recipes based on TDNNs [18], BLSTMs, and TDNN-LSTM hybrids with dropout [19, 11], and our TDNN-F recipe with all the additional features discussed above, including l2 regularization, “floating” semi-orthogonal constraint (Sec.”
Twin Regularization for online speech recognition (2018, background): “A popular choice is to feed the RNN with a context window that embeds some future frames [14, 15].”
From Senones to Chenones: Tied Context-Dependent Graphemes for Hybrid Speech Recognition (2019, background): “The first hidden layer of the LC-BLSTM subsamples the input by a factor of two [9], so that the posterior is emitted at a reduced 20ms frame rate.”
A Study of Enhancement, Augmentation, and Autoencoder Methods for Domain Adaptation in Distant Speech Recognition (2018, unknown): “Following [35], the context sizes of the TDNNs from layer one to seven are [−1, 0, 1], [−1, 0, 1], [−1, 0, 1], [−3, 0, 3], [−3, 0, 3], [−3, 0, 3], where [i, 0, k] indicates the summation of hidden vectors at time t+ i, t, and t+ k.”
Confidence Estimation and Deletion Prediction Using Bidirectional Recurrent Neural Networks (2018, methodology): “All speech recognisers made use of lattice-free maximum mutual information (LF-MMI) interleaved time-delay and long short-term memory neural network (TDNN-LSTM) [29, 30] acoustic models and n -gram language models [31].”
End-to-end Speech Recognition Using Lattice-free MMI (2018, methodology): “In all the experiments we use a TDNNLSTM structure [14], which has interleaving LSTM [29] and TDNN layers [30]; please refer to [14] for more details.”
Gpu-Accelerated Viterbi Exact Lattice Decoder for Batched Online and Offline Speech Recognition (2019, background): “Historically, the computational complexity of the acoustic model has dominated the inference execution time, and has been the focus of a variety of optimizations, including unusual network architectures, striding, and quantization techniques [4–6].”
Large-Scale Pre-Training of End-to-End Multi-Talker ASR for Meeting Transcription with Single Distant Microphone (2021, background): “While promising results were shown, all the previous studies were limited to either simulated data [16, 17, 18, 19, 20, 21, 22, 24, 25, 26, 27] or small-scale real data [7, 9, 8, 10](1).”
Scaling Up Online Speech Recognition Using ConvNets (2020, background): “While latency controlled bidirectional LSTMs are commonly used in online speech recognition [13], incorporating future context with convolutions can yield more accurate and lower latency models [14].”
On Training Recurrent Networks with Truncated Backpropagation Through time in Speech Recognition (2018, methodology): “Recurrent neural networks (RNNs) have been extensively used for speech and language processing, and are particularly successful at sequence prediction tasks, such as language modeling [1, 2, 3] and automatic speech recognition [4, 5, 6].”
Acoustic Modeling for Distant Multi-talker Speech Recognition with Single- and Multi-channel Branches (2019, background): “4 TDNN-BLSTM [16] Single BeamformIt 34.”