Unsupervised stress detection algorithm and experiments with real life data

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

6 Citations (Scopus)

Abstract

Stress is the major problem in the modern society and a reason for at least half of lost working days in European enterprises, but existing stress detectors are not sufficiently convenient for everyday use. One reason is that stress perception and stress manifestation vary a lot between individuals; hence, "one-fits-all-persons" stress detectors usually achieve notably lower accuracies than person-specific methods. The majority of existing approaches to person-specific stress recognition, however, employ fully supervised training, requiring to collect fairly large sets of labelled data from each end user. These sets should contain examples of stresses and normal conditions, and such data collection effort may be tiring for end users. Therefore this work proposes an algorithm to train person-specific stress detectors using only unlabelled data, not necessarily containing examples of stresses. The proposed method, based on Hidden Markov Models with maximum posterior marginal decision rule, was tested using real life data of 28 persons and achieved average stress detection accuracy of 75%, which is similar to the accuracies of state-of-the-art supervised algorithms for real life data.
Original languageEnglish
Title of host publicationProgress in Artificial Intelligence
EditorsZita Vale, Eugenio Oliveira, Joao Gama, Henrique Lopes Cardoso
PublisherSpringer
Pages95-107
Number of pages13
ISBN (Print)978-3-319-65339-6, 978-3-319-65340-2
DOIs
Publication statusPublished - 1 Jan 2017
MoE publication typeA4 Article in a conference publication
EventPortuguese Conference on Artificial Intelligence, EPIA 2017 - Porto, Portugal
Duration: 5 Sep 20178 Sep 2017

Publication series

SeriesLecture Notes in Computer Science
Volume10423 LNAI
ISSN0302-9743

Conference

ConferencePortuguese Conference on Artificial Intelligence, EPIA 2017
Abbreviated titleEPIA 2017
CountryPortugal
CityPorto
Period5/09/178/09/17

Fingerprint

Experiments
Detectors
Hidden Markov models
Industry

Keywords

  • stress detection
  • unsupervised learning
  • hidden Markov models

Cite this

Vildjiounaite, E., Kallio, J., Mäntyjärvi, J., Kyllönen, V., Lindholm, M., & Gimel'farb, G. (2017). Unsupervised stress detection algorithm and experiments with real life data. In Z. Vale, E. Oliveira, J. Gama, & H. Lopes Cardoso (Eds.), Progress in Artificial Intelligence (pp. 95-107). Springer. Lecture Notes in Computer Science, Vol.. 10423 LNAI https://doi.org/10.1007/978-3-319-65340-2_9
Vildjiounaite, Elena ; Kallio, Johanna ; Mäntyjärvi, Jani ; Kyllönen, Vesa ; Lindholm, Mikko ; Gimel'farb, Georgy. / Unsupervised stress detection algorithm and experiments with real life data. Progress in Artificial Intelligence. editor / Zita Vale ; Eugenio Oliveira ; Joao Gama ; Henrique Lopes Cardoso. Springer, 2017. pp. 95-107 (Lecture Notes in Computer Science, Vol. 10423 LNAI).
@inproceedings{85df0ff2d9414e358318f58dd3274ab6,
title = "Unsupervised stress detection algorithm and experiments with real life data",
abstract = "Stress is the major problem in the modern society and a reason for at least half of lost working days in European enterprises, but existing stress detectors are not sufficiently convenient for everyday use. One reason is that stress perception and stress manifestation vary a lot between individuals; hence, {"}one-fits-all-persons{"} stress detectors usually achieve notably lower accuracies than person-specific methods. The majority of existing approaches to person-specific stress recognition, however, employ fully supervised training, requiring to collect fairly large sets of labelled data from each end user. These sets should contain examples of stresses and normal conditions, and such data collection effort may be tiring for end users. Therefore this work proposes an algorithm to train person-specific stress detectors using only unlabelled data, not necessarily containing examples of stresses. The proposed method, based on Hidden Markov Models with maximum posterior marginal decision rule, was tested using real life data of 28 persons and achieved average stress detection accuracy of 75{\%}, which is similar to the accuracies of state-of-the-art supervised algorithms for real life data.",
keywords = "stress detection, unsupervised learning, hidden Markov models",
author = "Elena Vildjiounaite and Johanna Kallio and Jani M{\"a}ntyj{\"a}rvi and Vesa Kyll{\"o}nen and Mikko Lindholm and Georgy Gimel'farb",
note = "CA2: BA1614 CA2: BA16 CA2: BA1617 AU2: Vildjiounaite, Elena AU2: Kallio, Johanna AU2: M{\"a}ntyj{\"a}rvi, Jani AU2: Kyll{\"o}nen, Vesa AU2: Lindholm, Mikko",
year = "2017",
month = "1",
day = "1",
doi = "10.1007/978-3-319-65340-2_9",
language = "English",
isbn = "978-3-319-65339-6",
series = "Lecture Notes in Computer Science",
publisher = "Springer",
pages = "95--107",
editor = "Zita Vale and Eugenio Oliveira and Joao Gama and {Lopes Cardoso}, Henrique",
booktitle = "Progress in Artificial Intelligence",
address = "Germany",

}

Vildjiounaite, E, Kallio, J, Mäntyjärvi, J, Kyllönen, V, Lindholm, M & Gimel'farb, G 2017, Unsupervised stress detection algorithm and experiments with real life data. in Z Vale, E Oliveira, J Gama & H Lopes Cardoso (eds), Progress in Artificial Intelligence. Springer, Lecture Notes in Computer Science, vol. 10423 LNAI, pp. 95-107, Portuguese Conference on Artificial Intelligence, EPIA 2017, Porto, Portugal, 5/09/17. https://doi.org/10.1007/978-3-319-65340-2_9

Unsupervised stress detection algorithm and experiments with real life data. / Vildjiounaite, Elena; Kallio, Johanna; Mäntyjärvi, Jani; Kyllönen, Vesa; Lindholm, Mikko; Gimel'farb, Georgy.

Progress in Artificial Intelligence. ed. / Zita Vale; Eugenio Oliveira; Joao Gama; Henrique Lopes Cardoso. Springer, 2017. p. 95-107 (Lecture Notes in Computer Science, Vol. 10423 LNAI).

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

TY - GEN

T1 - Unsupervised stress detection algorithm and experiments with real life data

AU - Vildjiounaite, Elena

AU - Kallio, Johanna

AU - Mäntyjärvi, Jani

AU - Kyllönen, Vesa

AU - Lindholm, Mikko

AU - Gimel'farb, Georgy

N1 - CA2: BA1614 CA2: BA16 CA2: BA1617 AU2: Vildjiounaite, Elena AU2: Kallio, Johanna AU2: Mäntyjärvi, Jani AU2: Kyllönen, Vesa AU2: Lindholm, Mikko

PY - 2017/1/1

Y1 - 2017/1/1

N2 - Stress is the major problem in the modern society and a reason for at least half of lost working days in European enterprises, but existing stress detectors are not sufficiently convenient for everyday use. One reason is that stress perception and stress manifestation vary a lot between individuals; hence, "one-fits-all-persons" stress detectors usually achieve notably lower accuracies than person-specific methods. The majority of existing approaches to person-specific stress recognition, however, employ fully supervised training, requiring to collect fairly large sets of labelled data from each end user. These sets should contain examples of stresses and normal conditions, and such data collection effort may be tiring for end users. Therefore this work proposes an algorithm to train person-specific stress detectors using only unlabelled data, not necessarily containing examples of stresses. The proposed method, based on Hidden Markov Models with maximum posterior marginal decision rule, was tested using real life data of 28 persons and achieved average stress detection accuracy of 75%, which is similar to the accuracies of state-of-the-art supervised algorithms for real life data.

AB - Stress is the major problem in the modern society and a reason for at least half of lost working days in European enterprises, but existing stress detectors are not sufficiently convenient for everyday use. One reason is that stress perception and stress manifestation vary a lot between individuals; hence, "one-fits-all-persons" stress detectors usually achieve notably lower accuracies than person-specific methods. The majority of existing approaches to person-specific stress recognition, however, employ fully supervised training, requiring to collect fairly large sets of labelled data from each end user. These sets should contain examples of stresses and normal conditions, and such data collection effort may be tiring for end users. Therefore this work proposes an algorithm to train person-specific stress detectors using only unlabelled data, not necessarily containing examples of stresses. The proposed method, based on Hidden Markov Models with maximum posterior marginal decision rule, was tested using real life data of 28 persons and achieved average stress detection accuracy of 75%, which is similar to the accuracies of state-of-the-art supervised algorithms for real life data.

KW - stress detection

KW - unsupervised learning

KW - hidden Markov models

UR - http://www.scopus.com/inward/record.url?scp=85028981097&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-65340-2_9

DO - 10.1007/978-3-319-65340-2_9

M3 - Conference article in proceedings

SN - 978-3-319-65339-6

SN - 978-3-319-65340-2

T3 - Lecture Notes in Computer Science

SP - 95

EP - 107

BT - Progress in Artificial Intelligence

A2 - Vale, Zita

A2 - Oliveira, Eugenio

A2 - Gama, Joao

A2 - Lopes Cardoso, Henrique

PB - Springer

ER -

Vildjiounaite E, Kallio J, Mäntyjärvi J, Kyllönen V, Lindholm M, Gimel'farb G. Unsupervised stress detection algorithm and experiments with real life data. In Vale Z, Oliveira E, Gama J, Lopes Cardoso H, editors, Progress in Artificial Intelligence. Springer. 2017. p. 95-107. (Lecture Notes in Computer Science, Vol. 10423 LNAI). https://doi.org/10.1007/978-3-319-65340-2_9