Convolutional Neural Networks for Speech Recognition

By Ossama Abdel-Hamid, Abdel-rahman Mohamed, Hui Jiang, Li Deng, Gerald Penn, Dong Yu. 2014

2,165 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

Conformer: Convolution-augmented Transformer for Speech Recognition (2020, background): “Recurrent neural networks (RNNs) have been the defacto choice for ASR [1, 2, 3, 4] as they can model the temporal dependencies in the audio sequences effectively [5].”
Light Gated Recurrent Units for Speech Recognition (2018, background): “include, among the others, the development of effective regularization methods [2], [3], improved optimization algorithms [4], and better architectures [5]–[8].”
Fully Convolutional Speech Recognition (2018, methodology): “They have been first introduced as TDNNs to predict phoneme classes [10], and later to generate HMM posteriorgrams [11].”
Recent advances in convolutional neural networks (2015, neutral): “earn lters to process the raw waveform jointly with the rest of the network [311{315]. Most of the early applications of CNN in ASR only use fewer convolution layers. For example, Abdel-Hamid et al . [308] use one convolutional layer in their network, and Amodei et al . [316] use three convolutional”
Surrogate-Assisted Evolutionary Deep Learning Using an End-to-End Random Forest-Based Performance Predictor (2020, background): “CONVOLUTIONAL neural networks (CNNs) [1], [2], as the most dominant deep learning approaches [3], have been demonstrating their promising performance in addressing the various real-world applications, such as image classification [4], speech understanding [5], and natural language processing [6], to”
Recent Trends in Deep Learning Based Natural Language Processing (2017, methodology): “Speech recognition also requires such invariance and, thus, [60] used a hybrid CNN-HMM model which provided”
Residual LSTM: Design of a Deep Recurrent Architecture for Distant Speech Recognition (2017, background): “For example, convolutional neural networks (CNN) which has been huge success in image classification and detection were effective to reduce environmental and speaker variability in acoustic features [6, 7, 8, 9, 10].”
Leverage Financial News to Predict Stock Price Movements Using Word Embeddings and Deep Neural Networks (2015, background): “Convolutional neural network (CNN) [39] is the network with convolution and pooling structure that has been shown to be powerful in image processing tasks such as object detections.”
Transformer-Based Acoustic Modeling for Hybrid Speech Recognition (2019, background): “To address these issues, a variety of neural network architectures have been proposed to replace RNNs, including time delay neural networks (TDNN) [5], feed-forward sequential memory networks (FSMN) [6], and convolution neural networks (CNN) [4,13], while only limited success has been achieved.”
Permutation invariant training of deep models for speaker-independent multi-talker speech separation (2016, methodology): “as implemented using the computational network toolkit (CNTK) [8]. In all our experiments we used simple feed-forward DNNs with three hidden layers each with 1024 units, instead of more powerful CNNs [1, 17] and LSTMs. We show that even with this simple model, our approach can already perform surpri”
Between-Class Learning for Image Classification (2017, background): “Deep convolutional neural networks (CNNs) [18] have achieved high performance in various tasks, such as image recognition [17, 11], speech recognition [1, 21], and sound recognition [19, 26].”
End-To-End Source Separation With Adaptive Front-Ends (2017, background): “As shown in figure 2, the resulting spectral energies do not maintain locality [10] and exhibit a modulation that could be dependent on the frequency, the window size and the hop size parameters.”
Robust Spatial Filtering With Graph Convolutional Neural Networks (2017, background): “Convolutional Neural Networks (CNNs) have forever changed the pattern recognition landscape with breakthrough results on image classification [1]–[4], object detection [5], [6], and speech recognition [7].”
Convolutional Neural Networks for Medical Image Analysis: Full Training or Fine Tuning? (2016, methodology): “For instance, fine-tuning of an unsupervised model was used in [75] for acoustic speech recognition and in [76] for detection of epilepsy in EEG recordings.”
Learning Affective Features With a Hybrid Deep Model for Audio–Visual Emotion Recognition (2018, methodology): “To explain why we extract Mel-spectrogram segments with a length of 64 frames rather than 15 frames widely used in speech recognition [51], we compare the performance of two types of extracted Mel-spectrograms with different length, i.e. , 64 × 64 × 3 and 64 × 15 × 3.”
Blending LSTMs into CNNs (2015, methodology): “et al., 1998) were considered for speech for a long time (LeCun & Bengio, 1998; Lee et al., 2009), though only recently have become very successful (AbdelHamid et al., 2012; Sainath et al., 2013; Abdel-Hamid et al., 2014; Sainath et al., 2015a). These CNN architectures are quite different from t”
Deep-FSMN for Large Vocabulary Continuous Speech Recognition (2018, neutral): “Depending on how the networks are connected, there exist various types of deep neural networks, such as feedforward fully-connected neural networks(FNNs) [1, 2] , convolutional neural networks(CNNs)[3, 4] and recurrent neural networks (RNNs) [5, 6].”