Small energy masking for improved neural network training for end-to-end speech recognition
Small energy masking for improved neural network training for end-to-end speech recognition
About this item
Full title
Author / Creator
Publisher
Ithaca: Cornell University Library, arXiv.org
Journal title
Language
English
Formats
Publication information
Publisher
Ithaca: Cornell University Library, arXiv.org
Subjects
More information
Scope and Contents
Contents
In this paper, we present a Small Energy Masking (SEM) algorithm, which masks inputs having values below a certain threshold. More specifically, a time-frequency bin is masked if the filterbank energy in this bin is less than a certain energy threshold. A uniform distribution is employed to randomly generate the ratio of this energy threshold to th...
Alternative Titles
Full title
Small energy masking for improved neural network training for end-to-end speech recognition
Authors, Artists and Contributors
Author / Creator
Identifiers
Primary Identifiers
Record Identifier
TN_cdi_proquest_journals_2357104176
Permalink
https://devfeature-collection.sl.nsw.gov.au/record/TN_cdi_proquest_journals_2357104176
Other Identifiers
E-ISSN
2331-8422
