Open Speech and Language Resources


Identifier: SLR87

Summary: Chinese hotwords detection dataset, provided by Mobvoi CO.,LTD

Category: Speech

License: Apache License v.2.0

Downloads (use a mirror closer to you):
mobvoi_hotword_dataset.tgz [17G]   (Wave files of keyword and non-keyword data )   Mirrors: [US]   [EU]   [CN]  
mobvoi_hotword_dataset_resources.tgz [6.3M]   (Label, speaker and channel information of above wave files )   Mirrors: [US]   [EU]   [CN]  

About this resource:

The MobvoiHotwords is a corpus of wake-up words collected from a commercial smart speaker of Mobvoi. It consists of keyword and non-keyword utterances.

For keyword data, keyword utterances contain either 'Hi xiaowen' or 'Nihao Wenwen' are collected. For each keyword, there are about 36k utterances. All keyword data is collected from 788 subjects, ages 3-65, with different distances from the smart speaker (1, 3 and 5 meters). Different noises (typical home environment noises like music and TV) with varying sound pressure levels are played in the background during the collection. The keyword data is identical to the keyword data used in the paper below:

  author    = {Jingyong Hou and
               Yangyang Shi and
               Mari Ostendorf and
               Mei{-}Yuh Hwang and
               Lei Xie},
  title     = {Region Proposal Network Based Small-Footprint Keyword Spotting},
  journal   = {{IEEE} Signal Process. Lett.},
  volume    = {26},
  number    = {10},
  pages     = {1471--1475},
  year      = {2019},
  url       = {},
  doi       = {10.1109/LSP.2019.2936282}
There are also ~220 hours non-keyword data can be used as negative training samples, collected from the same smart speaker.