Voice activity detection in noisy conditions using tiny convolutional neural network
The paper investigates the problem of voice activity detection from a noisy sound signal. An extremely compact convolutional neural network is proposed. The model has only 385 trainable parameters. Proposed model doesn’t require a lot of computational resources that allows to use it as part of the “...
Main Authors: | , |
---|---|
Format: | Article |
Language: | Russian |
Published: |
The United Institute of Informatics Problems of the National Academy of Sciences of Belarus
2020-06-01
|
Series: | Informatika |
Subjects: | |
Online Access: | https://inf.grid.by/jour/article/view/968 |
id |
doaj-892d054342b7416282cf66bd60992042 |
---|---|
record_format |
Article |
spelling |
doaj-892d054342b7416282cf66bd609920422021-07-28T21:07:30ZrusThe United Institute of Informatics Problems of the National Academy of Sciences of Belarus Informatika1816-03012020-06-01172364310.37661/1816-0301-2020-17-2-36-43928Voice activity detection in noisy conditions using tiny convolutional neural networkR. S. Vashkevich0E. S. Azarov1Belarusian State University of Informatics and RadioelectronicsBelarusian State University of Informatics and RadioelectronicsThe paper investigates the problem of voice activity detection from a noisy sound signal. An extremely compact convolutional neural network is proposed. The model has only 385 trainable parameters. Proposed model doesn’t require a lot of computational resources that allows to use it as part of the “internet of things” concept for compact low power devices. At the same time the model provides state of the art results in voice activity detection in terms of detection accuracy. The properties of the model are achieved by using a special convolutional layer that considers the harmonic structure of vocal speech. This layer also eliminates redundancy of the model because it has invariance to changes of fundamental frequency. The model performance is evaluated in various noise conditions with different signal-to-noise ratios. The results show that the proposed model provides higher accuracy compared to voice activity detection model from the WebRTC framework by Google.https://inf.grid.by/jour/article/view/968voice activity detectorharmonic signalconvolutional neural networkpitchspeech processing |
collection |
DOAJ |
language |
Russian |
format |
Article |
sources |
DOAJ |
author |
R. S. Vashkevich E. S. Azarov |
spellingShingle |
R. S. Vashkevich E. S. Azarov Voice activity detection in noisy conditions using tiny convolutional neural network Informatika voice activity detector harmonic signal convolutional neural network pitch speech processing |
author_facet |
R. S. Vashkevich E. S. Azarov |
author_sort |
R. S. Vashkevich |
title |
Voice activity detection in noisy conditions using tiny convolutional neural network |
title_short |
Voice activity detection in noisy conditions using tiny convolutional neural network |
title_full |
Voice activity detection in noisy conditions using tiny convolutional neural network |
title_fullStr |
Voice activity detection in noisy conditions using tiny convolutional neural network |
title_full_unstemmed |
Voice activity detection in noisy conditions using tiny convolutional neural network |
title_sort |
voice activity detection in noisy conditions using tiny convolutional neural network |
publisher |
The United Institute of Informatics Problems of the National Academy of Sciences of Belarus |
series |
Informatika |
issn |
1816-0301 |
publishDate |
2020-06-01 |
description |
The paper investigates the problem of voice activity detection from a noisy sound signal. An extremely compact convolutional neural network is proposed. The model has only 385 trainable parameters. Proposed model doesn’t require a lot of computational resources that allows to use it as part of the “internet of things” concept for compact low power devices. At the same time the model provides state of the art results in voice activity detection in terms of detection accuracy. The properties of the model are achieved by using a special convolutional layer that considers the harmonic structure of vocal speech. This layer also eliminates redundancy of the model because it has invariance to changes of fundamental frequency. The model performance is evaluated in various noise conditions with different signal-to-noise ratios. The results show that the proposed model provides higher accuracy compared to voice activity detection model from the WebRTC framework by Google. |
topic |
voice activity detector harmonic signal convolutional neural network pitch speech processing |
url |
https://inf.grid.by/jour/article/view/968 |
work_keys_str_mv |
AT rsvashkevich voiceactivitydetectioninnoisyconditionsusingtinyconvolutionalneuralnetwork AT esazarov voiceactivitydetectioninnoisyconditionsusingtinyconvolutionalneuralnetwork |
_version_ |
1721262790768852992 |