Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2014 Jun 6;14(6):9995-10023.
doi: 10.3390/s140609995.

Dealing with the effects of sensor displacement in wearable activity recognition

Oresti Banos  1 , Mate Attila Toth  2 , Miguel Damas  3 , Hector Pomares  4 , Ignacio Rojas  5
Affiliations

Dealing with the effects of sensor displacement in wearable activity recognition

Oresti Banos et al. Sensors (Basel). .

Abstract

Most wearable activity recognition systems assume a predefined sensor deployment that remains unchanged during runtime. However, this assumption does not reflect real-life conditions. During the normal use of such systems, users may place the sensors in a position different from the predefined sensor placement. Also, sensors may move from their original location to a different one, due to a loose attachment. Activity recognition systems trained on activity patterns characteristic of a given sensor deployment may likely fail due to sensor displacements. In this work, we innovatively explore the effects of sensor displacement induced by both the intentional misplacement of sensors and self-placement by the user. The effects of sensor displacement are analyzed for standard activity recognition techniques, as well as for an alternate robust sensor fusion method proposed in a previous work. While classical recognition models show little tolerance to sensor displacement, the proposed method is proven to have notable capabilities to assimilate the changes introduced in the sensor position due to self-placement and provides considerable improvements for large misplacements.

PubMed Disclaimer

Figures

Figure 1.
Figure 1.
Example of sensor displacement introduced during the user self-placement of a sensor (a), and its effect at the feature level (b). In this particular example, the displacement with respect to the predefined deployment in the self-placement case applies to the right calf (RC), while the placement remains approximately similar for the sensor attached to the left calf (LC). In (b), the mean and standard deviation computed from the sensor acceleration signals is represented for various instances of a given activity. Confidence ellipses are plotted to represent the feature distribution for each case.
Figure 2.
Figure 2.
Example of possible sensor placements according to the (a) ideal, (b) self-placement and (c) induced-displacement deployments. In (b), the sensor is arbitrarily rotated 180° (approximately) by the user with respect to the ideal positioning (a). In (c), the expert explicitly displaces the sensor from the middle upper arm to the elbow.
Figure 3.
Figure 3.
Experimental setup (cardio-fitness room). Eight Xsens units are placed on each body limb and an additional one on the back. A laptop is used to store the recorded data and for labeling tasks. A camera records each session for offline post-processing. Sensor legend: left calf (LC), left thigh (LT), right calf (RC), right thigh (RT), back (BACK), left lower arm (LLA), left upper arm (LUA), right lower arm (RLA) and right upper arm (RUA).
Figure 4.
Figure 4.
Missing activity data for each particular subject (shading spots). (a) For ideal and self-placement conditions: the legend identifies the corresponding sensor deployment (both ≡ self-placed and ideally-placed). (b) For the induced-displaced condition: only Participants 2, 5 and 15 were considered.
Figure 5.
Figure 5.
Shading spots identify the displaced sensors for the (a) self-placement and (b) induced-displacement deployments. Only Participants 2, 5 and 15 were considered in (b).
Figure 6.
Figure 6.
The accuracy (average bar and standard deviation whiskers) results from the evaluation of the single sensor approach across all subjects and sensors for the (a) ideal-placement; (b) self-placement and (c) induced-displacement settings. The top legend identifies the classification paradigm. The horizontal axis labels identify the feature set used for each experiment. The activity recognition dataset (i.e., the number of activities) used is respectively underlined.
Figure 7.
Figure 7.
Accuracy (mean and standard deviation) results from the evaluation of the feature fusion model for the (a) ideal-placement, (b) self-placement and (c–f) induced-displacement (number of sensors) settings. The top legend identifies the classification paradigm. The horizontal axis labels identify the feature set used for each experiment. The activity recognition dataset (i.e., the number of activities) used is respectively underlined.
Figure 8.
Figure 8.
Accuracy (mean and standard deviation) results from the evaluation of the decision fusion model for the (a) ideal-placement; (b) self-placement and (c–f) induced-displacement (number of sensors) settings. The top legend identifies the base classifiers paradigm. The horizontal axis labels identify the feature set used for each experiment. The activity recognition dataset (i.e., the number of activities) used is respectively underlined.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Ermes M., Parkka J., Mantyjarvi J., Korhonen I. Detection of daily activities and sports with wearable sensors in controlled and uncontrolled conditions. IEEE Trans. Inf. Technol. Biomed. 2008;12:20–26. - PubMed
    1. Stiefmeier T., Roggen D., Ogris G., Lukowicz P., Tröster G. Wearable activity tracking in car manufacturing. IEEE Pervas. Comput. 2008;7:42–50.
    1. Chernbumroong S., Atkins A., Yu H. Activity classification using a single wrist-worn accelerometer. Proceedings of the 5th International Conference on Software Knowledge Information, Industrial Management and Applications; Benevento, Italy. 8–11 Spetember 2011; 2011. pp. 1–6.
    1. Mannini A., Intille S.S., Rosenberger M., Sabatini A.M., Haskell W. Activity recognition using a single accelerometer placed at the wrist or ankle. Med. Sci. Sports Exerc. 2013;45:2193–2203. - PMC - PubMed
    1. Atallah L., Lo B., King R., Yang G.Z. Sensorpositioning for activity recognition using wearable accelerometers. IEEE Trans. Biomed. Circuits Syst. 2011;5:320–329. - PubMed

Publication types