Small Footprint Convolutional Recurrent Networks for Streaming Wakeword Detection

In this work, we propose small footprint Convolutional Recurrent Neural Network models applied to the problem of wakeword detection and augment them with scaled dot product attention. We find that false accepts compared to Convolutional Neural Network models in a 250k parameter budget can be reduced...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2020-11
Main Authors: Mohammad Omar Khursheed, Christin, Jose, Kumar, Rajath, Fu, Gengshen, Kulis, Brian, Cheekatmalla, Santosh Kumar
Format: Article
Language:English
Subjects:
Artificial neural networks
Budgets
Mathematical models
Network latency
Neural networks
Parameters
Recurrent neural networks
Reduction
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In this work, we propose small footprint Convolutional Recurrent Neural Network models applied to the problem of wakeword detection and augment them with scaled dot product attention. We find that false accepts compared to Convolutional Neural Network models in a 250k parameter budget can be reduced by 25% with a 10% reduction in parameter size by using CRNNs, and we can get up to 32% improvement at a 50k parameter budget with 75% reduction in parameter size compared to word-level Dense Neural Network models. We discuss solutions to the challenging problem of performing inference on streaming audio with CRNNs, as well as differences in start-end index errors and latency in comparison to CNN, DNN, and DNN-HMM models.
ISSN:2331-8422