| Titre : | Development of a Web Application for Speech Recognition using Deep Recurrent Neural Networks | | Type de document : | projet fin études | | Auteurs : | ALAOUI MDARHRI / MAHDI REGAL SAAD, Auteur | | Langues : | Français (fre) | | Catégories : | Internet des Objets et Services Mobiles ( IOSM )
| | Mots-clés : | Automatic Speech Recognition(ASR), Deep learning, Machine Learning,HMM,
DTW, ANN, RNN, FFNN, LSTM, GRU. | | Index. décimale : | mast 225/19 | | Résumé : | Over the past few decades, there has been tremendous development in deep learning paradigms
used in automatic speech recognition (ASR). which is the very case in this work as we will be
focusing on de ning at rst speech as a digital amenable signal and di erent techniques used
to identify words and phrases in a spoken language. A comprehensive review of classical ma-
chine learning techniques like HMM, DTW, ANN models employed in ASR is also provided
along with a thorough review on the recent developments in deep learning and its variant
architectures(RNN, FFNN, LSTM, GRU) which has provided signi cant improvements in
ASR performance.
Later on we will be also discussing in details the general procedure that will be conducted in
order to develop and set a model for speech recognition based on recurrent neural networks.
To do so, we have created a platform for both web and mobile users using di erent APIs
(IBM, Google,Azure,Amazon) and also using a stand-alone neural networks model trained
on a chosen set of trigger words that have been collected through a gate which we have built
and made accessible to the public.
|
Development of a Web Application for Speech Recognition using Deep Recurrent Neural Networks [projet fin études] / ALAOUI MDARHRI / MAHDI REGAL SAAD, Auteur . - [s.d.]. Langues : Français ( fre) | Catégories : | Internet des Objets et Services Mobiles ( IOSM )
| | Mots-clés : | Automatic Speech Recognition(ASR), Deep learning, Machine Learning,HMM,
DTW, ANN, RNN, FFNN, LSTM, GRU. | | Index. décimale : | mast 225/19 | | Résumé : | Over the past few decades, there has been tremendous development in deep learning paradigms
used in automatic speech recognition (ASR). which is the very case in this work as we will be
focusing on de ning at rst speech as a digital amenable signal and di erent techniques used
to identify words and phrases in a spoken language. A comprehensive review of classical ma-
chine learning techniques like HMM, DTW, ANN models employed in ASR is also provided
along with a thorough review on the recent developments in deep learning and its variant
architectures(RNN, FFNN, LSTM, GRU) which has provided signi cant improvements in
ASR performance.
Later on we will be also discussing in details the general procedure that will be conducted in
order to develop and set a model for speech recognition based on recurrent neural networks.
To do so, we have created a platform for both web and mobile users using di erent APIs
(IBM, Google,Azure,Amazon) and also using a stand-alone neural networks model trained
on a chosen set of trigger words that have been collected through a gate which we have built
and made accessible to the public.
|
|
Faire une suggestion Affiner la recherche
Development of a Web Application for Speech Recognition using Deep Recurrent Neural Networks / ALAOUI MDARHRI / MAHDI REGAL SAAD
