Combining Clustering and Functionals based Acoustic Feature Representations for Classification of Baby Sounds

H. Kaya; Oxana Verkholyak; Maxim Markitantov; Alexey Karpov; Maxim Markitantov

doi:10.1145/3395035.3425182

Combining Clustering and Functionals based Acoustic Feature Representations for Classification of Baby Sounds

H. Kaya, Oxana Verkholyak, Maxim Markitantov, Alexey Karpov, Maxim Markitantov

Research output: Contribution to conference › Paper › Academic

Abstract

This paper investigates different fusion strategies as well as provides insights on their effectiveness alongside standalone classifiers in the framework of paralinguistic analysis of infant vocalizations. The combinations of such systems as Support Vector Machines (SVM) and Extreme Learning Machines (ELM) based classifiers, as well as its weighted kernel version are explored, training systems on different acoustic feature representations and implementing weighted score-level fusion of the predictions. The proposed framework is tested on INTERSPEECH ComParE-2019 Baby Sounds corpus, which is a collection of Home Bank infant vocalization corpora annotated for five classes. Adhering to the challenge protocol, using a single test set submission we outperform the challenge baseline Unweighted Average Recall (UAR) score and achieve a comparable result to the state-of-the-art.

Original language	English
Pages	509-513
Number of pages	5
DOIs	https://doi.org/10.1145/3395035.3425182
Publication status	Published - 29 Oct 2020
Event	ICMI 2020 Workshop on Bridging Social Sciences and AI for Understanding Child Behavior - Virtual Event, Utrecht, Netherlands Duration: 29 Oct 2020 → 29 Oct 2020 https://sites.google.com/view/wocbu/home

Workshop

Workshop	ICMI 2020 Workshop on Bridging Social Sciences and AI for Understanding Child Behavior
Abbreviated title	WoCBU
Country/Territory	Netherlands
City	Utrecht
Period	29/10/20 → 29/10/20
Internet address	https://sites.google.com/view/wocbu/home

Keywords

baby sounds classification
computational paralinguistics
information fusion
extreme learning machines
support vector machines

Access to Document

10.1145/3395035.3425182

Cite this

@conference{9d0339196d9e4bb5a3ea7ce60384958c,

title = "Combining Clustering and Functionals based Acoustic Feature Representations for Classification of Baby Sounds",

abstract = "This paper investigates different fusion strategies as well as provides insights on their effectiveness alongside standalone classifiers in the framework of paralinguistic analysis of infant vocalizations. The combinations of such systems as Support Vector Machines (SVM) and Extreme Learning Machines (ELM) based classifiers, as well as its weighted kernel version are explored, training systems on different acoustic feature representations and implementing weighted score-level fusion of the predictions. The proposed framework is tested on INTERSPEECH ComParE-2019 Baby Sounds corpus, which is a collection of Home Bank infant vocalization corpora annotated for five classes. Adhering to the challenge protocol, using a single test set submission we outperform the challenge baseline Unweighted Average Recall (UAR) score and achieve a comparable result to the state-of-the-art.",

keywords = "baby sounds classification, computational paralinguistics, information fusion, extreme learning machines, support vector machines",

author = "H. Kaya and Oxana Verkholyak and Maxim Markitantov and Alexey Karpov and Maxim Markitantov",

year = "2020",

month = oct,

day = "29",

doi = "10.1145/3395035.3425182",

language = "English",

pages = "509--513",

note = "ICMI 2020 Workshop on Bridging Social Sciences and AI for Understanding Child Behavior, WoCBU ; Conference date: 29-10-2020 Through 29-10-2020",

url = "https://sites.google.com/view/wocbu/home",

}

Kaya, H, Verkholyak, O, Markitantov, M, Karpov, A & Markitantov, M 2020, 'Combining Clustering and Functionals based Acoustic Feature Representations for Classification of Baby Sounds', Paper presented at ICMI 2020 Workshop on Bridging Social Sciences and AI for Understanding Child Behavior, Utrecht, Netherlands, 29/10/20 - 29/10/20 pp. 509-513. https://doi.org/10.1145/3395035.3425182

Combining Clustering and Functionals based Acoustic Feature Representations for Classification of Baby Sounds. / Kaya, H.; Verkholyak, Oxana; Markitantov, Maxim et al.
2020. 509-513 Paper presented at ICMI 2020 Workshop on Bridging Social Sciences and AI for Understanding Child Behavior, Utrecht, Netherlands.

Research output: Contribution to conference › Paper › Academic

TY - CONF

T1 - Combining Clustering and Functionals based Acoustic Feature Representations for Classification of Baby Sounds

AU - Kaya, H.

AU - Verkholyak, Oxana

AU - Markitantov, Maxim

AU - Karpov, Alexey

AU - Markitantov, Maxim

PY - 2020/10/29

Y1 - 2020/10/29

N2 - This paper investigates different fusion strategies as well as provides insights on their effectiveness alongside standalone classifiers in the framework of paralinguistic analysis of infant vocalizations. The combinations of such systems as Support Vector Machines (SVM) and Extreme Learning Machines (ELM) based classifiers, as well as its weighted kernel version are explored, training systems on different acoustic feature representations and implementing weighted score-level fusion of the predictions. The proposed framework is tested on INTERSPEECH ComParE-2019 Baby Sounds corpus, which is a collection of Home Bank infant vocalization corpora annotated for five classes. Adhering to the challenge protocol, using a single test set submission we outperform the challenge baseline Unweighted Average Recall (UAR) score and achieve a comparable result to the state-of-the-art.

AB - This paper investigates different fusion strategies as well as provides insights on their effectiveness alongside standalone classifiers in the framework of paralinguistic analysis of infant vocalizations. The combinations of such systems as Support Vector Machines (SVM) and Extreme Learning Machines (ELM) based classifiers, as well as its weighted kernel version are explored, training systems on different acoustic feature representations and implementing weighted score-level fusion of the predictions. The proposed framework is tested on INTERSPEECH ComParE-2019 Baby Sounds corpus, which is a collection of Home Bank infant vocalization corpora annotated for five classes. Adhering to the challenge protocol, using a single test set submission we outperform the challenge baseline Unweighted Average Recall (UAR) score and achieve a comparable result to the state-of-the-art.

KW - baby sounds classification

KW - computational paralinguistics

KW - information fusion

KW - extreme learning machines

KW - support vector machines

U2 - 10.1145/3395035.3425182

DO - 10.1145/3395035.3425182

M3 - Paper

SP - 509

EP - 513

T2 - ICMI 2020 Workshop on Bridging Social Sciences and AI for Understanding Child Behavior

Y2 - 29 October 2020 through 29 October 2020

ER -

Combining Clustering and Functionals based Acoustic Feature Representations for Classification of Baby Sounds

Abstract

Workshop

Keywords

Access to Document

Fingerprint

Cite this