Deep Residual Learning for Small-Footprint Keyword Spotting

Raphael Tang, Jimmy Lin
10/28/2017
cs.CL

Official code: castorini/honk (527 stars, MIT, last commit 2023-05-22).

Abstract

We explore the application of deep residual learning and dilated convolutions to the keyword spotting task, using the recently-released Google Speech Commands Dataset as our benchmark. Our best residual network (ResNet) implementation significantly outperforms Google's previous convolutional neural networks in terms of accuracy. By varying model depth and width, we can achieve compact models that also outperform previous small-footprint variants. To our knowledge, we are the first to examine these approaches for keyword spotting, and our results establish an open-source state-of-the-art reference to support the development of future speech-based interfaces.

View on arXivView PDF

Code Implementations(1)

castorini/honkOfficial100%
MIT

PyTorch implementations of neural network models for keyword spotting

527120Shell, PythonSep 14, 20173 years agoMIT

Similar Papers

Keyword spotting -- Detecting commands in speech using deep learning

87% match2023

Honk: A PyTorch Reimplementation of Convolutional Neural Networks for Keyword Spotting

86% match1 repo5272017

Vocal Tract Length Warped Features for Spoken Keyword Spotting

83% match2025

Efficient Continual Learning in Keyword Spotting using Binary Neural Networks

82% match2025

Hello Edge: Keyword Spotting on Microcontrollers

82% match1 repo1.3k2017

Keyword Transformer: A Self-Attention Model for Keyword Spotting

82% match1 repo1482021

Cite this paper

@article{tang2017deep,
  title  = {Deep Residual Learning for Small-Footprint Keyword Spotting},
  author = {Raphael Tang and Jimmy Lin},
  year   = {2017},
  eprint = {1710.10361},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1710.10361}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1710.10361.svg)](https://codeofpaper.com/paper/1710.10361)

Discussion