Keyword-Spotting-Playground

AI Research into Spoken Keyword Spotting.
Collection of PyTorch implementations of Spoken Keyword Spotting presented in research papers.
Model architectures will not always mirror the ones proposed in the papers, but I have chosen to focus on getting the core ideas covered instead of getting every layer configuration right.

Table of Contents

• Installation
• Implementations
  • Temporal Convolution Resnet(TC-ResNet)
  • Broadcasting Residual Network(BC-ResNet)

Implementations

About DataSet

Speech Commands DataSet is a set of one-second .wav audio files, each containing a single spoken English word.
These words are from a small set of commands, and are spoken by a variety of different speakers.
The audio files are organized into folders based on the word they contain, and this dataset is designed to help train simple machine learning models.

Installation

We use the Google Speech Commands Dataset (GSC) as the training data. By running the script, you can download the training data:

cd <ROOT>/dataset
bash download_gsc.sh

Temporal Convolution Resnet

Temporal Convolution for Real-time Keyword Spotting on Mobile Devices
[Paper] [Code]

Broadcasting Residual Network

Broadcasted Residual Learning for Efficient Keyword Spotting
[Paper] [Code]

Reference

1. https://github.com/hyperconnect/TC-ResNet
2. https://github.com/eriklindernoren/PyTorch-GAN
3. https://github.com/roman-vygon/BCResNet

Acknowledgement

Thanks for the contribution by huangyz0918!

GitHub

View Github