Evaluation of supervised classification algorithms for human activity recognition with inertial sensors

Tahmina Zebin; Patricia J. Scully; Krikor B. Ozanyan

doi:10.1109/ICSENS.2017.8234222

Evaluation of supervised classification algorithms for human activity recognition with inertial sensors

Tahmina Zebin, Patricia J. Scully, Krikor B. Ozanyan

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

30 Citations (Scopus)

29 Downloads (Pure)

Abstract

The main aim of this work is to compare the performance of different algorithms for human activity recognition by extracting various statistical time domain and frequency domain features from the inertial sensor data. Our results show that Support Vector Machines with quadratic kernel classifier (accuracy: 93.5%) and Ensemble classifier with bagging and boosting (accuracy: 94.6%) outperforms other known activity classification algorithms. A parallel coordinate plot based on visualization of features is used to identify useful features or predictors for separating classes. This enabled exclusion of features that contribute least to classification accuracy in a multi-sensor system (five in our case), made the classifier lightweight in terms of number of useful features, training time and computational load and lends itself to real-time implementation.

Original language	English
Title of host publication	IEEE SENSORS 2017 - Conference Proceedings
Publisher	The Institute of Electrical and Electronics Engineers (IEEE)
Pages	1-3
Number of pages	3
Volume	2017-December
ISBN (Electronic)	9781509010127
DOIs	https://doi.org/10.1109/ICSENS.2017.8234222
Publication status	Published - 21 Dec 2017
Externally published	Yes
Event	16th IEEE SENSORS Conference - Glasgow, United Kingdom Duration: 30 Oct 2017 → 1 Nov 2017

Conference

Conference	16th IEEE SENSORS Conference
Abbreviated title	ICSENS 2017
Country/Territory	United Kingdom
City	Glasgow
Period	30/10/17 → 1/11/17

Keywords

Classification
Feature Selection
Human Activity Recognition (HAR)
MATLAB
Supervised Machine Learning

Access to Document

10.1109/ICSENS.2017.8234222

Accepted_ManuscriptAccepted author manuscript, 864 KBLicence: Other

Cite this

@inproceedings{aa112cf7f7ce4eee8b7b84b474680ddb,

title = "Evaluation of supervised classification algorithms for human activity recognition with inertial sensors",

abstract = "The main aim of this work is to compare the performance of different algorithms for human activity recognition by extracting various statistical time domain and frequency domain features from the inertial sensor data. Our results show that Support Vector Machines with quadratic kernel classifier (accuracy: 93.5%) and Ensemble classifier with bagging and boosting (accuracy: 94.6%) outperforms other known activity classification algorithms. A parallel coordinate plot based on visualization of features is used to identify useful features or predictors for separating classes. This enabled exclusion of features that contribute least to classification accuracy in a multi-sensor system (five in our case), made the classifier lightweight in terms of number of useful features, training time and computational load and lends itself to real-time implementation.",

keywords = "Classification, Feature Selection, Human Activity Recognition (HAR), MATLAB, Supervised Machine Learning",

author = "Tahmina Zebin and Scully, {Patricia J.} and Ozanyan, {Krikor B.}",

year = "2017",

month = dec,

day = "21",

doi = "10.1109/ICSENS.2017.8234222",

language = "English",

volume = "2017-December",

pages = "1--3",

booktitle = "IEEE SENSORS 2017 - Conference Proceedings",

publisher = "The Institute of Electrical and Electronics Engineers (IEEE)",

address = "United States",

note = "16th IEEE SENSORS Conference, ICSENS 2017 ; Conference date: 30-10-2017 Through 01-11-2017",

}

Zebin, T, Scully, PJ & Ozanyan, KB 2017, Evaluation of supervised classification algorithms for human activity recognition with inertial sensors. in IEEE SENSORS 2017 - Conference Proceedings. vol. 2017-December, The Institute of Electrical and Electronics Engineers (IEEE), pp. 1-3, 16th IEEE SENSORS Conference, Glasgow, United Kingdom, 30/10/17. https://doi.org/10.1109/ICSENS.2017.8234222

Evaluation of supervised classification algorithms for human activity recognition with inertial sensors. / Zebin, Tahmina; Scully, Patricia J.; Ozanyan, Krikor B.

IEEE SENSORS 2017 - Conference Proceedings. Vol. 2017-December The Institute of Electrical and Electronics Engineers (IEEE), 2017. p. 1-3.

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

TY - GEN

T1 - Evaluation of supervised classification algorithms for human activity recognition with inertial sensors

AU - Zebin, Tahmina

AU - Scully, Patricia J.

AU - Ozanyan, Krikor B.

PY - 2017/12/21

Y1 - 2017/12/21

N2 - The main aim of this work is to compare the performance of different algorithms for human activity recognition by extracting various statistical time domain and frequency domain features from the inertial sensor data. Our results show that Support Vector Machines with quadratic kernel classifier (accuracy: 93.5%) and Ensemble classifier with bagging and boosting (accuracy: 94.6%) outperforms other known activity classification algorithms. A parallel coordinate plot based on visualization of features is used to identify useful features or predictors for separating classes. This enabled exclusion of features that contribute least to classification accuracy in a multi-sensor system (five in our case), made the classifier lightweight in terms of number of useful features, training time and computational load and lends itself to real-time implementation.

AB - The main aim of this work is to compare the performance of different algorithms for human activity recognition by extracting various statistical time domain and frequency domain features from the inertial sensor data. Our results show that Support Vector Machines with quadratic kernel classifier (accuracy: 93.5%) and Ensemble classifier with bagging and boosting (accuracy: 94.6%) outperforms other known activity classification algorithms. A parallel coordinate plot based on visualization of features is used to identify useful features or predictors for separating classes. This enabled exclusion of features that contribute least to classification accuracy in a multi-sensor system (five in our case), made the classifier lightweight in terms of number of useful features, training time and computational load and lends itself to real-time implementation.

KW - Classification

KW - Feature Selection

KW - Human Activity Recognition (HAR)

KW - MATLAB

KW - Supervised Machine Learning

UR - http://www.scopus.com/inward/record.url?scp=85044278282&partnerID=8YFLogxK

U2 - 10.1109/ICSENS.2017.8234222

DO - 10.1109/ICSENS.2017.8234222

M3 - Conference contribution

AN - SCOPUS:85044278282

VL - 2017-December

SP - 1

EP - 3

BT - IEEE SENSORS 2017 - Conference Proceedings

PB - The Institute of Electrical and Electronics Engineers (IEEE)

T2 - 16th IEEE SENSORS Conference

Y2 - 30 October 2017 through 1 November 2017

ER -

Evaluation of supervised classification algorithms for human activity recognition with inertial sensors

Abstract

Conference

Keywords

Access to Document

Other files and links

Profiles

Tahmina Zebin

Cite this