The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
La reconnaissance d'activité à partir de capteurs est un problème de classification des données de séries chronologiques. Certaines recherches dans ce domaine utilisent des fonctionnalités artisanales dans les domaines temporel et fréquentiel qui diffèrent selon les ensembles de données. Une autre approche catégoriquement différente consiste à utiliser des méthodes d'apprentissage en profondeur pour l'apprentissage des fonctionnalités. Cet article explore un terrain d'entente dans lequel un extracteur de caractéristiques standard est utilisé pour générer un grand nombre de caractéristiques candidates dans le domaine temporel, suivi d'un sélecteur de caractéristiques conçu pour réduire le biais en faveur de techniques de classification spécifiques. De plus, cet article préconise l'utilisation de fonctionnalités qui sont pour la plupart insensibles à l'orientation du capteur et montre leur applicabilité au problème de reconnaissance d'activité. L'approche proposée est évaluée à l'aide de six ensembles de données différents accessibles au public, collectés dans diverses conditions en utilisant différents protocoles expérimentaux et montre une précision comparable ou supérieure à celle des méthodes de pointe sur la plupart des ensembles de données, mais en utilisant généralement un ordre de grandeur moins de fonctionnalités.
Yasser MOHAMMAD
AIST,Assiut University
Kazunori MATSUMOTO
KDDI Research Inc.
Keiichiro HOASHI
KDDI Research Inc.
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copier
Yasser MOHAMMAD, Kazunori MATSUMOTO, Keiichiro HOASHI, "Selecting Orientation-Insensitive Features for Activity Recognition from Accelerometers" in IEICE TRANSACTIONS on Information,
vol. E102-D, no. 1, pp. 104-115, January 2019, doi: 10.1587/transinf.2018EDP7092.
Abstract: Activity recognition from sensors is a classification problem over time-series data. Some research in the area utilize time and frequency domain handcrafted features that differ between datasets. Another categorically different approach is to use deep learning methods for feature learning. This paper explores a middle ground in which an off-the-shelf feature extractor is used to generate a large number of candidate time-domain features followed by a feature selector that was designed to reduce the bias toward specific classification techniques. Moreover, this paper advocates the use of features that are mostly insensitive to sensor orientation and show their applicability to the activity recognition problem. The proposed approach is evaluated using six different publicly available datasets collected under various conditions using different experimental protocols and shows comparable or higher accuracy than state-of-the-art methods on most datasets but usually using an order of magnitude fewer features.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2018EDP7092/_p
Copier
@ARTICLE{e102-d_1_104,
author={Yasser MOHAMMAD, Kazunori MATSUMOTO, Keiichiro HOASHI, },
journal={IEICE TRANSACTIONS on Information},
title={Selecting Orientation-Insensitive Features for Activity Recognition from Accelerometers},
year={2019},
volume={E102-D},
number={1},
pages={104-115},
abstract={Activity recognition from sensors is a classification problem over time-series data. Some research in the area utilize time and frequency domain handcrafted features that differ between datasets. Another categorically different approach is to use deep learning methods for feature learning. This paper explores a middle ground in which an off-the-shelf feature extractor is used to generate a large number of candidate time-domain features followed by a feature selector that was designed to reduce the bias toward specific classification techniques. Moreover, this paper advocates the use of features that are mostly insensitive to sensor orientation and show their applicability to the activity recognition problem. The proposed approach is evaluated using six different publicly available datasets collected under various conditions using different experimental protocols and shows comparable or higher accuracy than state-of-the-art methods on most datasets but usually using an order of magnitude fewer features.},
keywords={},
doi={10.1587/transinf.2018EDP7092},
ISSN={1745-1361},
month={January},}
Copier
TY - JOUR
TI - Selecting Orientation-Insensitive Features for Activity Recognition from Accelerometers
T2 - IEICE TRANSACTIONS on Information
SP - 104
EP - 115
AU - Yasser MOHAMMAD
AU - Kazunori MATSUMOTO
AU - Keiichiro HOASHI
PY - 2019
DO - 10.1587/transinf.2018EDP7092
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E102-D
IS - 1
JA - IEICE TRANSACTIONS on Information
Y1 - January 2019
AB - Activity recognition from sensors is a classification problem over time-series data. Some research in the area utilize time and frequency domain handcrafted features that differ between datasets. Another categorically different approach is to use deep learning methods for feature learning. This paper explores a middle ground in which an off-the-shelf feature extractor is used to generate a large number of candidate time-domain features followed by a feature selector that was designed to reduce the bias toward specific classification techniques. Moreover, this paper advocates the use of features that are mostly insensitive to sensor orientation and show their applicability to the activity recognition problem. The proposed approach is evaluated using six different publicly available datasets collected under various conditions using different experimental protocols and shows comparable or higher accuracy than state-of-the-art methods on most datasets but usually using an order of magnitude fewer features.
ER -