An Application of Recurrent Neural Networks to Discriminative Keyword Spotting

By Santiago Fernández, Alex Graves, J. Urgen Schmidhuber. 2007

274 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

Deep learning in neural networks: An overview (2014, background): “2013-: more contests and benchmark records A stack (Fernandez et al., 2007b; Graves & Schmidhuber, 2009) (Section 5.10) of bi-directional LSTMRNNs (Graves& Schmidhuber, 2005) trained by CTC (Sections 5.13, 5.17) broke a famous TIMIT speech (phoneme) recognition record, achieving 17.7% test set…”
Fast and accurate recurrent neural network acoustic models for speech recognition (2015, methodology): “There have been previous studies using LSTM RNN CTC models for keyword spotting tasks with small vocabularies (e.g. 12 words [24]).”
One Big Net For Everything (2018, background): “In the consolidation phase of Step 4, ONE then may rather easily learn [10, 50] the speech command-dependence of these behaviors through gradient-based learning, without having to interact with the environment again.”
Fruit recognition from images using deep learning (2017, background): “Traditionally, recurrent networks have been used to process sequential data, handwriting or speech recognition being the most known examples [13, 10].”
Applications for deep learning in ecology (2018, methodology): “A popular implementation of RNN is the Long Term Short-Memory network (LSTM), an architecture capable of learning long-term dependencies that has proven especially efficient for tasks such as speech recognition 70 or translation 71 .”
A Siamese Long Short-Term Memory Architecture for Human Re-identification (2016, background): “ reduced and more discriminative features can be learned. A variant of RNNs called Long Short-Term Memory (LSTM) [22] cells have been used to spot salient keywords in sentences [41] and speech inputs [13] to learn context (i.e., topic) relevant information. The advanced gating mechanisms inside the ”
Efficient Keyword Spotting Using Dilated Convolutions and Gating (2018, background): “Recent works have suggested RNN based keyword spotting using LSTM cells that can leverage longer temporal context using gating mechanism and internal states [7, 8, 9].”
Supervised Sequence Labelling with Recurrent Neural Networks (2012, methodology): “The following experiments confirm this observation by recording the word error rate both with and without the requirement that the network find the approximate location of the keywords (Fernández et al., 2007).”
A review on the long short-term memory model (2020, background): “Keyword spotting is a detection task consisting in discovering the presence of specific spoken words in speech signals (Fernández et al. 2007).”
Streaming small-footprint keyword spotting using sequence-to-sequence models (2017, background): “Many previous works propose to train neural networks to identify word targets in individual keywords: for example, using feed-forward deep neural networks [4, 5, 6], convolutional networks [7] or recurrent neural networks [8, 9, 10].”
Hello Edge: Keyword Spotting on Microcontrollers (2017, methodology): “Other techniques explored for KWS include discriminative models adopting a large-margin problem formulation [13] or recurrent neural networks (RNN) [14].”
Video-based emotion recognition using CNN-RNN and C3D hybrid networks (2016, background): “Recently, LSTMs are also well developed and shown to be efficient to deal with various sequences versus sequences problems, such as audio analysis [26, 29], video captioning [24, 25], video action recognition [22], and so on.”
Small-footprint keyword spotting using deep neural networks (2014, background): “Other recent work explores discriminative models for keyword spotting based on largemargin formulation [10, 11] or recurrent neural networks [12, 13].”
Speech Processing for Digital Home Assistants: Combining signal processing with deep-learning techniques (2019, methodology): “To improve detection accuracy, convolutional [47], timedelay [48], or recurrent network layers [49] have been proposed, as well as subword modeling of the keyword and background speech using a DNN-HMM architecture [50], all of which aim to exploit the temporal properties of the input signal for clas”
From speech to letters - using a novel neural network architecture for grapheme based ASR (2009, methodology): “The novel approach is evaluated extensively on the Wall Street Journal 1 corpus.”