Exploring Filterbank Learning for Keyword Spotting

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

1 Citation (Scopus)

Abstract

Despite their great performance over the years, handcrafted speech features are not necessarily optimal for any particular speech application. Consequently, with greater or lesser success, optimal filterbank learning has been studied for different speech processing tasks. In this paper, we fill in a gap by exploring filterbank learning for keyword spotting (KWS). Two approaches are examined: filterbank matrix learning in the power spectral domain and parameter learning of a psychoacoustically-motivated gammachirp filterbank. Filterbank parameters are optimized jointly with a modern deep residual neural network-based KWS back-end. Our experimental results reveal that, in general, there are no statistically significant differences, in terms of KWS accuracy, between using a learned filterbank and handcrafted speech features. Thus, while we conclude that the latter are still a wise choice when using modern KWS back-ends, we also hypothesize that this could be a symptom of information redundancy, which opens up new research possibilities in the field of small-footprint KWS.

Original languageEnglish
Title of host publication28th European Signal Processing Conference (EUSIPCO)
Number of pages5
PublisherIEEE
Publication date2021
Pages331-335
Article number9287772
ISBN (Print)978-1-7281-5001-7
ISBN (Electronic)978-9-0827-9705-3
DOIs
Publication statusPublished - 2021
Event2020 28th European Signal Processing Conference (EUSIPCO) - Amsterdam, Netherlands
Duration: 18 Jan 202121 Jan 2021

Conference

Conference2020 28th European Signal Processing Conference (EUSIPCO)
Country/TerritoryNetherlands
CityAmsterdam
Period18/01/202121/01/2021
SeriesProceedings of the European Signal Processing Conference
ISSN2076-1465

Keywords

  • End-to-end
  • Filterbank learning
  • Gammachirp filterbank
  • Gammatone filterbank
  • Keyword spotting

Fingerprint

Dive into the research topics of 'Exploring Filterbank Learning for Keyword Spotting'. Together they form a unique fingerprint.

Cite this