Survey on Deep Neural Networks in Speech and Vision Systems

Alam, Mahbubul; Samad, Manar D.; Vidyaratne, Lasitha; Glandon, Alexander; Iftekharuddin, Khan M.

doi:10.1016/j.neucom.2020.07.053

Cited by 184 publications

(67 citation statements)

References 153 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The following aspects describe the differences between end-to-end systems and traditional systems, although they may not necessarily be considered as fundamental differences. Some end-to-end considered neural networks, as the one in Section 3.5, which is introduced in [13], are capable of processing raw audio signal, the feature extraction being integrated into the network [14], but also some hand-crafted features which have been already extracted in a previous step. From the acoustic model point of view, while the traditional systems are modeling phonemes, the end-to-end systems are trained on graphemes (characters) [15], word-pieces [16] or even entire words [17].…”

Section: End-to-end Asrmentioning

confidence: 99%

Performance vs. hardware requirements in state-of-the-art automatic speech recognition

Georgescu

Pappalardo

Cucu

et al. 2021

J AUDIO SPEECH MUSIC PROC.

View full text Add to dashboard Cite

The last decade brought significant advances in automatic speech recognition (ASR) thanks to the evolution of deep learning methods. ASR systems evolved from pipeline-based systems, that modeled hand-crafted speech features with probabilistic frameworks and generated phone posteriors, to end-to-end (E2E) systems, that translate the raw waveform directly into words using one deep neural network (DNN). The transcription accuracy greatly increased, leading to ASR technology being integrated into many commercial applications. However, few of the existing ASR technologies are suitable for integration in embedded applications, due to their hard constrains related to computing power and memory usage. This overview paper serves as a guided tour through the recent literature on speech recognition and compares the most popular ASR implementations. The comparison emphasizes the trade-off between ASR performance and hardware requirements, to further serve decision makers in choosing the system which fits best their embedded application. To the best of our knowledge, this is the first study to provide this kind of trade-off analysis for state-of-the-art ASR systems.

show abstract

Section: End-to-end Asrmentioning

confidence: 99%

Performance vs. hardware requirements in state-of-the-art automatic speech recognition

Georgescu

Pappalardo

Cucu

et al. 2021

J AUDIO SPEECH MUSIC PROC.

View full text Add to dashboard Cite

show abstract

“…Nosúltimos anos, esforços significativos de pesquisa foram feitos em técnicas de deep neural network [Mallmann et al 2020], que atualmente fornece os resultados mais promissores em campos como classificação de imagens e detecção de objetos [Alam et al 2020]. Uma dessas arquiteturas de deep neural são os deep autoencoders, que normalmente são usados para redução de dimensionalidade [Li et al 2020], e eliminação de ruído de imagem [Gondara 2016], através de um processo de duas etapas, o codificador e o decodificador.…”

Section: Deep Autoencodersunclassified

Um modelo de detecção de intrusão baseado em deep autoencoders e transfer learning

Santos

Viegas

Santin

2021

Anais Do XXI Simpósio Brasileiro De Segurança Da Informação E De Sistemas Computacionais (SBSeg 2021)

View full text Add to dashboard Cite

As técnicas de aprendizado de máquina para detecção de intrusão baseada na rede geralmente pressupõem que o tráfego da rede não muda com o tempo ou que as atualizações do modelo podem ser realizadas facilmente. Neste artigo, propomos um novo modelo de detecção de intrusão baseado em deep autoencoders e transfer learning para facilitar a atualização do modelo. Experimentos realizados mostraram que as abordagens na literatura são incapazes de lidar com mudanças de tráfego de rede ao longo do tempo. A abordagem proposta é capaz de melhorar a taxa de falso positivo em até 23,9%, e fornecer taxas de precisão semelhantes às técnicas tradicionais, exigindo apenas 22% de dados de treinamento e 28% dos custos computacionais.

show abstract

“…Deep learning is a class of ML that allows multiple hidden layers for data processing [ 19 ]. A deep neural network (DNN) combines an artificial neural network with deep learning and is capable of providing a better solution to problems in cognitive learning such as speech and image recognition [ 20 ]. So far, DNN models have been successfully applied to learn and predict a range of properties of diverse types of materials, including metals, ceramics, and macromolecular materials [ 21 , 22 , 23 ].…”

Section: Introductionmentioning

confidence: 99%

A Deep Neural Network for Accurate and Robust Prediction of the Glass Transition Temperature of Polyhydroxyalkanoate Homo- and Copolymers

Jiang

Marrone

et al. 2020

Materials

View full text Add to dashboard Cite

The purpose of this study was to develop a data-driven machine learning model to predict the performance properties of polyhydroxyalkanoates (PHAs), a group of biosourced polyesters featuring excellent performance, to guide future design and synthesis experiments. A deep neural network (DNN) machine learning model was built for predicting the glass transition temperature, Tg, of PHA homo- and copolymers. Molecular fingerprints were used to capture the structural and atomic information of PHA monomers. The other input variables included the molecular weight, the polydispersity index, and the percentage of each monomer in the homo- and copolymers. The results indicate that the DNN model achieves high accuracy in estimation of the glass transition temperature of PHAs. In addition, the symmetry of the DNN model is ensured by incorporating symmetry data in the training process. The DNN model achieved better performance than the support vector machine (SVD), a nonlinear ML model and least absolute shrinkage and selection operator (LASSO), a sparse linear regression model. The relative importance of factors affecting the DNN model prediction were analyzed. Sensitivity of the DNN model, including strategies to deal with missing data, were also investigated. Compared with commonly used machine learning models incorporating quantitative structure–property (QSPR) relationships, it does not require an explicit descriptor selection step but shows a comparable performance. The machine learning model framework can be readily extended to predict other properties.

show abstract

Survey on Deep Neural Networks in Speech and Vision Systems

Cited by 184 publications

References 153 publications

Performance vs. hardware requirements in state-of-the-art automatic speech recognition

Performance vs. hardware requirements in state-of-the-art automatic speech recognition

Um modelo de detecção de intrusão baseado em deep autoencoders e transfer learning

A Deep Neural Network for Accurate and Robust Prediction of the Glass Transition Temperature of Polyhydroxyalkanoate Homo- and Copolymers

Contact Info

Product

Resources

About