Masked Conditional Neural Networks for Environmental Sound Classification

Fady Medhat; David Chesmore; John Robinson

doi:10.1007/978-3-319-71078-5_2

Masked Conditional Neural Networks for Environmental Sound Classification

Fady Medhat, David Chesmore, John Robinson

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

The ConditionaL Neural Network (CLNN) exploits the nature of the temporal sequencing of the sound signal represented in a spectrogram, and its variant the Masked ConditionaL Neural Network (MCLNN) induces the network to learn in frequency bands by embedding a filterbank-like sparseness over the network’s links using a binary mask. Additionally, the masking automates the exploration of different feature combinations concurrently analogous to handcrafting the optimum combination of features for a recognition task. We have evaluated the MCLNN performance using the Urbansound8k dataset of environmental sounds. Additionally, we present a collection of manually recorded sounds for rail and road traffic, YorNoise, to investigate the confusion rates among machine generated sounds possessing low-frequency components. MCLNN has achieved competitive results without augmentation and using 12% of the trainable parameters utilized by an equivalent model based on state-of-the-art Convolutional Neural Networks on the Urbansound8k. We extended the Urbansound8k dataset with YorNoise, where experiments have shown that common tonal properties affect the classification performance.

Original language	English
Title of host publication	Artificial Intelligence XXXIV
Subtitle of host publication	37th SGAI International Conference on Artificial Intelligence, AI 2017, Cambridge, UK, December 12-14, 2017, Proceedings
Publisher	Springer
Pages	21-33
ISBN (Electronic)	9783319710785
ISBN (Print)	9783319710778
DOIs	https://doi.org/10.1007/978-3-319-71078-5_2
Publication status	Published - Feb 2018
Event	Artificial Intelligence XXXIV - , United Kingdom Duration: 12 Dec 2017 → …

Publication series

Name	Lecture Notes in Computer Science
Publisher	Springer
Volume	10630

Conference

Conference	Artificial Intelligence XXXIV
Abbreviated title	SGAI
Country/Territory	United Kingdom
Period	12/12/17 → …

Access to Document

10.1007/978-3-319-71078-5_2

https://arxiv.org/abs/1805.10004

Cite this

Medhat, F., Chesmore, D., & Robinson, J. (2018). Masked Conditional Neural Networks for Environmental Sound Classification. In Artificial Intelligence XXXIV: 37th SGAI International Conference on Artificial Intelligence, AI 2017, Cambridge, UK, December 12-14, 2017, Proceedings (pp. 21-33). (Lecture Notes in Computer Science ; Vol. 10630). Springer. https://doi.org/10.1007/978-3-319-71078-5_2

@inproceedings{2d6a8709f9ee4365a9ef031608873006,

title = "Masked Conditional Neural Networks for Environmental Sound Classification",

abstract = "The ConditionaL Neural Network (CLNN) exploits the nature of the temporal sequencing of the sound signal represented in a spectrogram, and its variant the Masked ConditionaL Neural Network (MCLNN) induces the network to learn in frequency bands by embedding a filterbank-like sparseness over the network{\textquoteright}s links using a binary mask. Additionally, the masking automates the exploration of different feature combinations concurrently analogous to handcrafting the optimum combination of features for a recognition task. We have evaluated the MCLNN performance using the Urbansound8k dataset of environmental sounds. Additionally, we present a collection of manually recorded sounds for rail and road traffic, YorNoise, to investigate the confusion rates among machine generated sounds possessing low-frequency components. MCLNN has achieved competitive results without augmentation and using 12% of the trainable parameters utilized by an equivalent model based on state-of-the-art Convolutional Neural Networks on the Urbansound8k. We extended the Urbansound8k dataset with YorNoise, where experiments have shown that common tonal properties affect the classification performance.",

author = "Fady Medhat and David Chesmore and John Robinson",

year = "2018",

month = feb,

doi = "10.1007/978-3-319-71078-5_2",

language = "English",

isbn = "9783319710778",

series = "Lecture Notes in Computer Science ",

publisher = "Springer",

pages = "21--33",

booktitle = "Artificial Intelligence XXXIV",

address = "Germany",

note = "Artificial Intelligence XXXIV, SGAI ; Conference date: 12-12-2017",

}

Medhat, F , Chesmore, D & Robinson, J 2018, Masked Conditional Neural Networks for Environmental Sound Classification. in Artificial Intelligence XXXIV: 37th SGAI International Conference on Artificial Intelligence, AI 2017, Cambridge, UK, December 12-14, 2017, Proceedings. Lecture Notes in Computer Science , vol. 10630, Springer, pp. 21-33, Artificial Intelligence XXXIV, United Kingdom, 12/12/17. https://doi.org/10.1007/978-3-319-71078-5_2

Masked Conditional Neural Networks for Environmental Sound Classification. / Medhat, Fady ; Chesmore, David ; Robinson, John.
Artificial Intelligence XXXIV: 37th SGAI International Conference on Artificial Intelligence, AI 2017, Cambridge, UK, December 12-14, 2017, Proceedings. Springer, 2018. p. 21-33 (Lecture Notes in Computer Science ; Vol. 10630).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Masked Conditional Neural Networks for Environmental Sound Classification

AU - Medhat, Fady

AU - Chesmore, David

AU - Robinson, John

PY - 2018/2

Y1 - 2018/2

N2 - The ConditionaL Neural Network (CLNN) exploits the nature of the temporal sequencing of the sound signal represented in a spectrogram, and its variant the Masked ConditionaL Neural Network (MCLNN) induces the network to learn in frequency bands by embedding a filterbank-like sparseness over the network’s links using a binary mask. Additionally, the masking automates the exploration of different feature combinations concurrently analogous to handcrafting the optimum combination of features for a recognition task. We have evaluated the MCLNN performance using the Urbansound8k dataset of environmental sounds. Additionally, we present a collection of manually recorded sounds for rail and road traffic, YorNoise, to investigate the confusion rates among machine generated sounds possessing low-frequency components. MCLNN has achieved competitive results without augmentation and using 12% of the trainable parameters utilized by an equivalent model based on state-of-the-art Convolutional Neural Networks on the Urbansound8k. We extended the Urbansound8k dataset with YorNoise, where experiments have shown that common tonal properties affect the classification performance.

AB - The ConditionaL Neural Network (CLNN) exploits the nature of the temporal sequencing of the sound signal represented in a spectrogram, and its variant the Masked ConditionaL Neural Network (MCLNN) induces the network to learn in frequency bands by embedding a filterbank-like sparseness over the network’s links using a binary mask. Additionally, the masking automates the exploration of different feature combinations concurrently analogous to handcrafting the optimum combination of features for a recognition task. We have evaluated the MCLNN performance using the Urbansound8k dataset of environmental sounds. Additionally, we present a collection of manually recorded sounds for rail and road traffic, YorNoise, to investigate the confusion rates among machine generated sounds possessing low-frequency components. MCLNN has achieved competitive results without augmentation and using 12% of the trainable parameters utilized by an equivalent model based on state-of-the-art Convolutional Neural Networks on the Urbansound8k. We extended the Urbansound8k dataset with YorNoise, where experiments have shown that common tonal properties affect the classification performance.

U2 - 10.1007/978-3-319-71078-5_2

DO - 10.1007/978-3-319-71078-5_2

M3 - Conference contribution

SN - 9783319710778

T3 - Lecture Notes in Computer Science

SP - 21

EP - 33

BT - Artificial Intelligence XXXIV

PB - Springer

T2 - Artificial Intelligence XXXIV

Y2 - 12 December 2017

ER -