Structural sparsification for far-field speaker recognition with intel R GNA

Conference Paper

Recently, deep neural networks (DNN) have been widely used in speaker recognition area. In order to achieve fast response time and high accuracy, the requirements for hardware resources increase rapidly. However, as the speaker recognition application is often implemented on mobile devices, it is necessary to maintain a low computational cost while keeping high accuracy in far-field condition. In this paper, we apply structural sparsification on time-delay neural networks (TDNN) to remove redundant structures and accelerate the execution. On our targeted hardware, our model can remove 60% of parameters and only slightly increasing equal error rate (EER) by 0.18% while our structural sparse model can achieve more than 1.5× speedup.

Full Text

Duke Authors

Cited Authors

  • Zhang, J; Huang, J; Deisher, M; Li, H; Chen, Y

Published Date

  • May 1, 2020

Published In

Volume / Issue

  • 2020-May /

Start / End Page

  • 3037 - 3041

International Standard Serial Number (ISSN)

  • 1520-6149

International Standard Book Number 13 (ISBN-13)

  • 9781509066315

Digital Object Identifier (DOI)

  • 10.1109/ICASSP40776.2020.9054569

Citation Source

  • Scopus