CCA based feature selection with application to continuous depression recognition from acoustic speech features

Heysem Kaya; Florian Eyben; Albert Ali Salah; Bjorn Schuller

doi:10.1109/ICASSP.2014.6854298

CCA based feature selection with application to continuous depression recognition from acoustic speech features

Heysem Kaya^*
, Florian Eyben
, Albert Ali Salah
, Bjorn Schuller

^*Corresponding author for this work

Technical University of Munich
Imperial College London

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

Abstract

In this study we make use of Canonical Correlation Analysis (CCA) based feature selection for continuous depression recognition from speech. Besides its common use in multi-modal/multi-view feature extraction, CCA can be easily employed as a feature selector. We introduce several novel ways of CCA based filter (ranking) methods, showing their relations to previous work. We test the suitability of proposed methods on the AVEC 2013 dataset under the ACM MM 2013 Challenge protocol. Using 17% of features, we obtained a relative improvement of 30% on the challenge's test-set baseline Root Mean Square Error.

Original language	English
Title of host publication	2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014
Publisher	IEEE
Pages	3729-3733
Number of pages	5
ISBN (Print)	9781479928927
DOIs	https://doi.org/10.1109/ICASSP.2014.6854298
Publication status	Published - 1 Jan 2014
Event	2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014 - Florence, Italy Duration: 4 May 2014 → 9 May 2014

Conference

Conference	2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014
Country/Territory	Italy
City	Florence
Period	4/05/14 → 9/05/14

Keywords

acoustic speech processing
affect recognition
Canonical Correlation Analysis
depression recognition
feature extraction
feature selection

Access to Document

10.1109/ICASSP.2014.6854298

Cite this

@inproceedings{078c49b830e54bfea81eabddf2073fc0,

title = "CCA based feature selection with application to continuous depression recognition from acoustic speech features",

abstract = "In this study we make use of Canonical Correlation Analysis (CCA) based feature selection for continuous depression recognition from speech. Besides its common use in multi-modal/multi-view feature extraction, CCA can be easily employed as a feature selector. We introduce several novel ways of CCA based filter (ranking) methods, showing their relations to previous work. We test the suitability of proposed methods on the AVEC 2013 dataset under the ACM MM 2013 Challenge protocol. Using 17\% of features, we obtained a relative improvement of 30\% on the challenge's test-set baseline Root Mean Square Error.",

keywords = "acoustic speech processing, affect recognition, Canonical Correlation Analysis, depression recognition, feature extraction, feature selection",

author = "Heysem Kaya and Florian Eyben and Salah, \{Albert Ali\} and Bjorn Schuller",

year = "2014",

month = jan,

day = "1",

doi = "10.1109/ICASSP.2014.6854298",

language = "English",

isbn = "9781479928927",

pages = "3729--3733",

booktitle = "2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014",

publisher = "IEEE",

address = "United States",

note = "2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014 ; Conference date: 04-05-2014 Through 09-05-2014",

}

Kaya, H, Eyben, F, Salah, AA & Schuller, B 2014, CCA based feature selection with application to continuous depression recognition from acoustic speech features. in 2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014., 6854298, IEEE, pp. 3729-3733, 2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014, Florence, Italy, 4/05/14. https://doi.org/10.1109/ICASSP.2014.6854298

CCA based feature selection with application to continuous depression recognition from acoustic speech features. / Kaya, Heysem; Eyben, Florian; Salah, Albert Ali et al.
2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014. IEEE, 2014. p. 3729-3733 6854298.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - CCA based feature selection with application to continuous depression recognition from acoustic speech features

AU - Kaya, Heysem

AU - Eyben, Florian

AU - Salah, Albert Ali

AU - Schuller, Bjorn

PY - 2014/1/1

Y1 - 2014/1/1

N2 - In this study we make use of Canonical Correlation Analysis (CCA) based feature selection for continuous depression recognition from speech. Besides its common use in multi-modal/multi-view feature extraction, CCA can be easily employed as a feature selector. We introduce several novel ways of CCA based filter (ranking) methods, showing their relations to previous work. We test the suitability of proposed methods on the AVEC 2013 dataset under the ACM MM 2013 Challenge protocol. Using 17% of features, we obtained a relative improvement of 30% on the challenge's test-set baseline Root Mean Square Error.

AB - In this study we make use of Canonical Correlation Analysis (CCA) based feature selection for continuous depression recognition from speech. Besides its common use in multi-modal/multi-view feature extraction, CCA can be easily employed as a feature selector. We introduce several novel ways of CCA based filter (ranking) methods, showing their relations to previous work. We test the suitability of proposed methods on the AVEC 2013 dataset under the ACM MM 2013 Challenge protocol. Using 17% of features, we obtained a relative improvement of 30% on the challenge's test-set baseline Root Mean Square Error.

KW - acoustic speech processing

KW - affect recognition

KW - Canonical Correlation Analysis

KW - depression recognition

KW - feature extraction

KW - feature selection

UR - https://www.scopus.com/pages/publications/84905226582

U2 - 10.1109/ICASSP.2014.6854298

DO - 10.1109/ICASSP.2014.6854298

M3 - Conference contribution

AN - SCOPUS:84905226582

SN - 9781479928927

SP - 3729

EP - 3733

BT - 2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014

PB - IEEE

T2 - 2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014

Y2 - 4 May 2014 through 9 May 2014

ER -

CCA based feature selection with application to continuous depression recognition from acoustic speech features

Abstract

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this