keyword-spotting-recognition-through-state-of-art-neural-architectures's Introduction

Keyword-Spotting-Recognition-through-state-of-art-neural-architectures

In this repo is presented the work by Francesco Ferretto and Federica Latora on the study of Deep Learning applications for Audio Speech Recognition (ASR).

It follows the tree structure of the project:

├── data
│   ├── binaries		<- .npy data
│   ├── processed		<- processed data (features)
│   └── raw			<- raw unedited data
│       ├── data_documentation	<- original data details
│       ├── dataset_v2_reduced  <- reduced dataset (10 classes)
│       │     ├── down
│       │     ├── go
│       │     ├── left
│       │     ├── no
│       │     ├── off
│       │     ├── on
│       │     ├── right
│       │     ├── stop
│       │     ├── up
│       │     └── yes   
│       └── dataset_v2  <- complete dataset (35 classes)
│             ├── backward
│             ├── bed
│             ├── bird
│             ├── cat
│             ├── dog
│             ├── down
│             ├── eight
│             ├── five
│             ├── follow
│             ├── forward
│             ├── four
│             ├── go
│             ├── happy
│             ├── house
│             ├── learn
│             ├── left
│             ├── marvin
│             ├── nine
│             ├── no
│             ├── off
│             ├── on
│             ├── one
│             ├── right
│             ├── seven
│             ├── sheila
│             ├── six
│             ├── stop
│             ├── three
│             ├── tree
│             ├── two
│             ├── up
│             ├── visual
│             ├── wow
│             ├── yes
│             └── zero
├── documents		    <- folder containing dictionaries and list of processed audio files during feature and pre-processing phases
├── models		    	<- folder containing the models in .h5 format
├── notebooks			<- prototyping notebooks for EDA, feature selection, experimental purposes for modeling (comparison, ensemble, ...)
└── src				    <- folder containing project source code
    ├── data			<- scripts to download/generate data
    ├── features		<- scripts for feature extraction
    └── model			<- scripts for training, evaluation and prediction

Recommend Projects