Efficient Keyword Spotting Using Dilated Convolutions and Gating
Authors:
- 1 May 2019
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 6351-6355
Abstract
We explore the application of end-to-end stateless temporal modeling to small-footprint keyword spotting as opposed to recurrent networks that model long-term temporal dependencies using internal states. We propose a model inspired by the recent success of dilated convolutions in sequence modeling applications, allowing to train deeper architectures in resource-constrained configurations. Gated activations and residual connections are also added, following a similar configuration to WaveNet. In addition, we apply a custom target labeling that back-propagates loss from specific frames of interest, therefore yielding higher accuracy and only requiring to detect the end of the keyword. Our experimental results show that our model outperforms a max-pooling loss trained recurrent neural network using LSTM cells, with a significant decrease in false rejection rate. The underlying dataset - "Hey Snips" utterances recorded by over 2.2K different speakers - has been made publicly available to establish an open reference for wake-word detection.Keywords
All Related Versions
This publication has 7 references indexed in Scilit:
- Temporal Modeling Using Dilated Convolution and Gating for Voice-Activity-Detection2018
- Max-pooling loss training of long short-term memory networks for small-footprint keyword spotting2017
- Online word-spotting in continuous speech with recurrent neural networks2014
- Small-footprint keyword spotting using deep neural networks2014
- A hidden Markov model based keyword recognition system2002
- Improvements and applications for key word recognition using hidden Markov modeling techniques1991
- Automatic recognition of keywords in unconstrained speech using hidden Markov modelsIEEE Transactions on Acoustics, Speech, and Signal Processing, 1990