Logo image
Back
A Lightweight Dynamic Filter For Keyword Spotting
Conference proceeding   Open access

A Lightweight Dynamic Filter For Keyword Spotting

Donghyeon Kim, Kyungdeuk Ko, Jeonggi Kwak, David K. Han and Hanseok Ko
2023 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW), pp 1-5
04 Jun 2023
DOI: https://doi.org/10.1109/ICASSPW59220.2023.10193258
url
http://arxiv.org/abs/2109.11165View
Submitted Open arXiv.org - Non-exclusive license to distribute

Abstract

Acoustics Adaptation models computational cost Computational modeling Convolution dynamic filter dynamic weight Feature extraction keyword spotting Speech recognition Training data
Keyword Spotting (KWS) from speech signals is widely applied to perform fully hands-free speech recognition. The KWS network is designed as a small-footprint model so it can continuously be active. Recent efforts have explored dynamic filter-based models in deep learning frameworks to enhance the system's robustness or accuracy. However, as a dynamic filter framework requires high computational costs, the implementation is limited to the computational condition of the device. In this paper, we propose a lightweight dynamic filter to improve the performance of KWS. Our proposed model divides the dynamic filter into two branches to reduce computational complexity: pixel level and instance level. The proposed lightweight dynamic filter is applied to the front end of KWS to enhance the separability of the input data. The experimental results show that our model is robustly working on unseen noise and small training data environments by using a small computational resource.

Metrics

10 Record Views
5 citations in Web of Science
2 citations in Scopus

Details

InCites Highlights

Data related to this publication, from InCites Benchmarking & Analytics tool:

Collaboration types
Domestic collaboration
International collaboration
Web of Science research areas
Acoustics
Computer Science, Interdisciplinary Applications
Engineering, Electrical & Electronic
Imaging Science & Photographic Technology
Logo image