RPN_KWS

Pytorch code of paper "Region Proposal Network Based Small-Footprint Keyword Spotting" https://ieeexplore.ieee.org/document/8807313

Please cite the work below if you want to use the code or want to do research related to our work

@ARTICLE{hou2019region, 
author={Hou, Jingyong and Shi, Yangyang and Ostendorf, Mari and  Hwang, Mei-Yuh
and Xie, Lei }, 
journal={IEEE Signal Processing Letters}, 
title={Region Proposal Network Based Small-Footprint Keyword Spotting}, 
year={2019}, 
volume={26}, 
number={10}, 
pages={1471-1475}
}

I will release a new version of RPN KWS with an Online Hard Example Mining (OHEM) algorithm, which will improve our system.

https://github.com/jingyonghou/RPN_KWS_OHEM

Detection samples

Selected two utterances which contains predefined keyword. The red box is the ground-truth start-end area of keyword from forced-alignment, the blue box is the best anchor selected according to the classification score, the green box is the proposed region proposal corresponding to the best anchor.

Running environment

Python 2.7.15

pytorch 0.4.1

CUDA 8.0 or higher

Kaldi

You should know basic knowledge of Kaldi before looking at the run script. I use Kaldi to extract Fbank features and do a global CMVN using the statictics from all training set. You should add cmd.sh, path.sh, steps and utils to your working dir before you run the script.

Please follow the run_rpn_kws.sh script to learn how to run the code

reference

https://github.com/jwyang/faster-rcnn.pytorch

https://github.com/vesis84/kaldi-io-for-python