Whodunnit - Searching for the most important feature types signalling emotion-related user states in speech

Anton Batliner; Stefan Steidl; Björn Schuller; Dino Seppi; Thurid Vogt; Johannes Wagner; Laurence Devillers; Laurence Vidrascu; Vered Aharonson; Loic Kessous; Noam Amir

doi:https://doi.org/10.1016/j.csl.2009.12.003

Whodunnit - Searching for the most important feature types signalling emotion-related user states in speech

Anton Batliner, Stefan Steidl, Björn Schuller, Dino Seppi, Thurid Vogt, Johannes Wagner, Laurence Devillers, Laurence Vidrascu, Vered Aharonson, Loic Kessous, Noam Amir

Department of Communication Disorders

Tel Aviv University

Research output: Contribution to journal › Article › peer-review

Abstract

In this article, we describe and interpret a set of acoustic and linguistic features that characterise emotional/emotion-related user states - confined to the one database processed: four classes in a German corpus of children interacting with a pet robot. To this end, we collected a very large feature vector consisting of more than 4000 features extracted at different sites. We performed extensive feature selection (Sequential Forward Floating Search) for seven acoustic and four linguistic types of features, ending up in a small number of 'most important' features which we try to interpret by discussing the impact of different feature and extraction types. We establish different measures of impact and discuss the mutual influence of acoustics and linguistics.

Original language	English
Pages (from-to)	4-28
Number of pages	25
Journal	Computer Speech and Language
Volume	25
Issue number	1
DOIs	https://doi.org/10.1016/j.csl.2009.12.003
State	Published - Jan 2011

Keywords

Automatic classification
Emotion
Feature selection
Feature types

All Science Journal Classification (ASJC) codes

Software
Theoretical Computer Science
Human-Computer Interaction

Access to Document

https://doi.org/10.1016/j.csl.2009.12.003

Cite this

@article{e8a09165df58442393d1cd297062249c,

title = "Whodunnit - Searching for the most important feature types signalling emotion-related user states in speech",

abstract = "In this article, we describe and interpret a set of acoustic and linguistic features that characterise emotional/emotion-related user states - confined to the one database processed: four classes in a German corpus of children interacting with a pet robot. To this end, we collected a very large feature vector consisting of more than 4000 features extracted at different sites. We performed extensive feature selection (Sequential Forward Floating Search) for seven acoustic and four linguistic types of features, ending up in a small number of 'most important' features which we try to interpret by discussing the impact of different feature and extraction types. We establish different measures of impact and discuss the mutual influence of acoustics and linguistics.",

keywords = "Automatic classification, Emotion, Feature selection, Feature types",

author = "Anton Batliner and Stefan Steidl and Bj{\"o}rn Schuller and Dino Seppi and Thurid Vogt and Johannes Wagner and Laurence Devillers and Laurence Vidrascu and Vered Aharonson and Loic Kessous and Noam Amir",

note = "Funding Information: The initiative to co-operate was taken within the European Network of Excellence (NoE) HUMAINE under the name Combining Efforts for Improving automatic Classification of Emotional user States (CEICES). This work was partly funded by the EU in the projects PF-STAR under Grant IST-2001-37599 and HUMAINE under Grant IST-2002-50742 . The responsibility lies with the authors.",

year = "2011",

month = jan,

doi = "https://doi.org/10.1016/j.csl.2009.12.003",

language = "الإنجليزيّة",

volume = "25",

pages = "4--28",

journal = "Computer Speech and Language",

issn = "0885-2308",

publisher = "Academic Press",

number = "1",

}

TY - JOUR

T1 - Whodunnit - Searching for the most important feature types signalling emotion-related user states in speech

AU - Batliner, Anton

AU - Steidl, Stefan

AU - Schuller, Björn

AU - Seppi, Dino

AU - Vogt, Thurid

AU - Wagner, Johannes

AU - Devillers, Laurence

AU - Vidrascu, Laurence

AU - Aharonson, Vered

AU - Kessous, Loic

AU - Amir, Noam

N1 - Funding Information: The initiative to co-operate was taken within the European Network of Excellence (NoE) HUMAINE under the name Combining Efforts for Improving automatic Classification of Emotional user States (CEICES). This work was partly funded by the EU in the projects PF-STAR under Grant IST-2001-37599 and HUMAINE under Grant IST-2002-50742 . The responsibility lies with the authors.

PY - 2011/1

Y1 - 2011/1

N2 - In this article, we describe and interpret a set of acoustic and linguistic features that characterise emotional/emotion-related user states - confined to the one database processed: four classes in a German corpus of children interacting with a pet robot. To this end, we collected a very large feature vector consisting of more than 4000 features extracted at different sites. We performed extensive feature selection (Sequential Forward Floating Search) for seven acoustic and four linguistic types of features, ending up in a small number of 'most important' features which we try to interpret by discussing the impact of different feature and extraction types. We establish different measures of impact and discuss the mutual influence of acoustics and linguistics.

AB - In this article, we describe and interpret a set of acoustic and linguistic features that characterise emotional/emotion-related user states - confined to the one database processed: four classes in a German corpus of children interacting with a pet robot. To this end, we collected a very large feature vector consisting of more than 4000 features extracted at different sites. We performed extensive feature selection (Sequential Forward Floating Search) for seven acoustic and four linguistic types of features, ending up in a small number of 'most important' features which we try to interpret by discussing the impact of different feature and extraction types. We establish different measures of impact and discuss the mutual influence of acoustics and linguistics.

KW - Automatic classification

KW - Emotion

KW - Feature selection

KW - Feature types

UR - http://www.scopus.com/inward/record.url?scp=77955412961&partnerID=8YFLogxK

U2 - https://doi.org/10.1016/j.csl.2009.12.003

DO - https://doi.org/10.1016/j.csl.2009.12.003

M3 - مقالة

SN - 0885-2308

VL - 25

SP - 4

EP - 28

JO - Computer Speech and Language

JF - Computer Speech and Language

IS - 1

ER -

Whodunnit - Searching for the most important feature types signalling emotion-related user states in speech

Abstract

Keywords

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this