Differentiating Between Walking and Stair Climbing Using Raw Accelerometry Data

William F. Fadel, Jacek K. Urbanek, Steven R. Albertson, Xiaochun Li, Andrea K. Chomistek, Jaroslaw Harezlak

Research output: Contribution to journalArticle

Abstract

Wearable accelerometers provide an objective measure of human physical activity. They record high-frequency unlabeled three-dimensional time series data. We extract meaningful features from the raw accelerometry data and based on them develop and evaluate a classification method for the detection of walking and its subclasses, i.e., level walking, descending stairs, and ascending stairs. Our methodology is tested on a sample of 32 middle-aged subjects for whom we extracted features based on the Fourier and wavelet transforms. We build subject-specific and group-level classification models utilizing a tree-based methodology. We evaluate the effects of sensor location and tuning parameters on the classification accuracy of the tree models. In the group-level classification setting, we propose a robust feature inter-subject normalization and evaluate its performance compared to unnormalized data. The overall classification accuracy for the three activities at the subject-specific level was on average 87.6%, with the ankle-worn accelerometers showing the best performance with an average accuracy 90.5%. At the group-level, the average overall classification accuracy for the three activities using the normalized features was 80.2% compared to 72.3% for the unnormalized features. In summary, a framework is provided for better use and feature extraction from raw accelerometry data to differentiate among different walking modalities as well as considerations for study design.

Original languageEnglish (US)
Pages (from-to)334-354
Number of pages21
JournalStatistics in Biosciences
Volume11
Issue number2
DOIs
StatePublished - Jul 15 2019

Fingerprint

Accelerometry
Stairs
Walking
Accelerometer
Accelerometers
Evaluate
Wavelet Analysis
Methodology
Location Parameter
Fourier Analysis
Time Series Data
Differentiate
Human Activities
Ankle
Modality
Wavelet transforms
Wavelet Transform
Feature Extraction
Normalization
Stair Climbing

Keywords

  • Accelerometer
  • Classification trees
  • Physical activity
  • Signal processing
  • Walking

ASJC Scopus subject areas

  • Statistics and Probability
  • Biochemistry, Genetics and Molecular Biology (miscellaneous)

Cite this

Differentiating Between Walking and Stair Climbing Using Raw Accelerometry Data. / Fadel, William F.; Urbanek, Jacek K.; Albertson, Steven R.; Li, Xiaochun; Chomistek, Andrea K.; Harezlak, Jaroslaw.

In: Statistics in Biosciences, Vol. 11, No. 2, 15.07.2019, p. 334-354.

Research output: Contribution to journalArticle

Fadel, William F. ; Urbanek, Jacek K. ; Albertson, Steven R. ; Li, Xiaochun ; Chomistek, Andrea K. ; Harezlak, Jaroslaw. / Differentiating Between Walking and Stair Climbing Using Raw Accelerometry Data. In: Statistics in Biosciences. 2019 ; Vol. 11, No. 2. pp. 334-354.
@article{e564c5a176104709ac03f781585d88fc,
title = "Differentiating Between Walking and Stair Climbing Using Raw Accelerometry Data",
abstract = "Wearable accelerometers provide an objective measure of human physical activity. They record high-frequency unlabeled three-dimensional time series data. We extract meaningful features from the raw accelerometry data and based on them develop and evaluate a classification method for the detection of walking and its subclasses, i.e., level walking, descending stairs, and ascending stairs. Our methodology is tested on a sample of 32 middle-aged subjects for whom we extracted features based on the Fourier and wavelet transforms. We build subject-specific and group-level classification models utilizing a tree-based methodology. We evaluate the effects of sensor location and tuning parameters on the classification accuracy of the tree models. In the group-level classification setting, we propose a robust feature inter-subject normalization and evaluate its performance compared to unnormalized data. The overall classification accuracy for the three activities at the subject-specific level was on average 87.6{\%}, with the ankle-worn accelerometers showing the best performance with an average accuracy 90.5{\%}. At the group-level, the average overall classification accuracy for the three activities using the normalized features was 80.2{\%} compared to 72.3{\%} for the unnormalized features. In summary, a framework is provided for better use and feature extraction from raw accelerometry data to differentiate among different walking modalities as well as considerations for study design.",
keywords = "Accelerometer, Classification trees, Physical activity, Signal processing, Walking",
author = "Fadel, {William F.} and Urbanek, {Jacek K.} and Albertson, {Steven R.} and Xiaochun Li and Chomistek, {Andrea K.} and Jaroslaw Harezlak",
year = "2019",
month = "7",
day = "15",
doi = "10.1007/s12561-019-09241-7",
language = "English (US)",
volume = "11",
pages = "334--354",
journal = "Statistics in Biosciences",
issn = "1867-1764",
publisher = "Springer New York",
number = "2",

}

TY - JOUR

T1 - Differentiating Between Walking and Stair Climbing Using Raw Accelerometry Data

AU - Fadel, William F.

AU - Urbanek, Jacek K.

AU - Albertson, Steven R.

AU - Li, Xiaochun

AU - Chomistek, Andrea K.

AU - Harezlak, Jaroslaw

PY - 2019/7/15

Y1 - 2019/7/15

N2 - Wearable accelerometers provide an objective measure of human physical activity. They record high-frequency unlabeled three-dimensional time series data. We extract meaningful features from the raw accelerometry data and based on them develop and evaluate a classification method for the detection of walking and its subclasses, i.e., level walking, descending stairs, and ascending stairs. Our methodology is tested on a sample of 32 middle-aged subjects for whom we extracted features based on the Fourier and wavelet transforms. We build subject-specific and group-level classification models utilizing a tree-based methodology. We evaluate the effects of sensor location and tuning parameters on the classification accuracy of the tree models. In the group-level classification setting, we propose a robust feature inter-subject normalization and evaluate its performance compared to unnormalized data. The overall classification accuracy for the three activities at the subject-specific level was on average 87.6%, with the ankle-worn accelerometers showing the best performance with an average accuracy 90.5%. At the group-level, the average overall classification accuracy for the three activities using the normalized features was 80.2% compared to 72.3% for the unnormalized features. In summary, a framework is provided for better use and feature extraction from raw accelerometry data to differentiate among different walking modalities as well as considerations for study design.

AB - Wearable accelerometers provide an objective measure of human physical activity. They record high-frequency unlabeled three-dimensional time series data. We extract meaningful features from the raw accelerometry data and based on them develop and evaluate a classification method for the detection of walking and its subclasses, i.e., level walking, descending stairs, and ascending stairs. Our methodology is tested on a sample of 32 middle-aged subjects for whom we extracted features based on the Fourier and wavelet transforms. We build subject-specific and group-level classification models utilizing a tree-based methodology. We evaluate the effects of sensor location and tuning parameters on the classification accuracy of the tree models. In the group-level classification setting, we propose a robust feature inter-subject normalization and evaluate its performance compared to unnormalized data. The overall classification accuracy for the three activities at the subject-specific level was on average 87.6%, with the ankle-worn accelerometers showing the best performance with an average accuracy 90.5%. At the group-level, the average overall classification accuracy for the three activities using the normalized features was 80.2% compared to 72.3% for the unnormalized features. In summary, a framework is provided for better use and feature extraction from raw accelerometry data to differentiate among different walking modalities as well as considerations for study design.

KW - Accelerometer

KW - Classification trees

KW - Physical activity

KW - Signal processing

KW - Walking

UR - http://www.scopus.com/inward/record.url?scp=85073651311&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85073651311&partnerID=8YFLogxK

U2 - 10.1007/s12561-019-09241-7

DO - 10.1007/s12561-019-09241-7

M3 - Article

AN - SCOPUS:85073651311

VL - 11

SP - 334

EP - 354

JO - Statistics in Biosciences

JF - Statistics in Biosciences

SN - 1867-1764

IS - 2

ER -