Multitaper mel-spectrograms for keyword spotting

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: de Souza, Douglas Baptista, Bakri, Khaled Jamal, Ferreira, Fernanda, Inacio, Juliana
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911945859268608
author de Souza, Douglas Baptista
Bakri, Khaled Jamal
Ferreira, Fernanda
Inacio, Juliana
author_facet de Souza, Douglas Baptista
Bakri, Khaled Jamal
Ferreira, Fernanda
Inacio, Juliana
contents Keyword spotting (KWS) is one of the speech recognition tasks most sensitive to the quality of the feature representation. However, the research on KWS has traditionally focused on new model topologies, putting little emphasis on other aspects like feature extraction. This paper investigates the use of the multitaper technique to create improved features for KWS. The experimental study is carried out for different test scenarios, windows and parameters, datasets, and neural networks commonly used in embedded KWS applications. Experiment results confirm the advantages of using the proposed improved features.
format Preprint
id arxiv_https___arxiv_org_abs_2407_04662
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Multitaper mel-spectrograms for keyword spotting
de Souza, Douglas Baptista
Bakri, Khaled Jamal
Ferreira, Fernanda
Inacio, Juliana
Audio and Speech Processing
Machine Learning
Keyword spotting (KWS) is one of the speech recognition tasks most sensitive to the quality of the feature representation. However, the research on KWS has traditionally focused on new model topologies, putting little emphasis on other aspects like feature extraction. This paper investigates the use of the multitaper technique to create improved features for KWS. The experimental study is carried out for different test scenarios, windows and parameters, datasets, and neural networks commonly used in embedded KWS applications. Experiment results confirm the advantages of using the proposed improved features.
title Multitaper mel-spectrograms for keyword spotting
topic Audio and Speech Processing
Machine Learning
url https://arxiv.org/abs/2407.04662