Achieving Accelerometer Wrist and Orientation Invariance in Physical Activity Classification via Domain Adaption

in Journal for the Measurement of Physical Behaviour
Restricted access

Purchase article

USD $24.95

Student 1 year subscription

USD $37.00

1 year subscription

USD $50.00

Student 2 year subscription

USD $71.00

2 year subscription

USD $93.00

Purpose: Physical activity classifiers are typically trained on data obtained from sensors at a set orientation. Changes in this orientation (such as being on a different wrist) result in performance degradation. This work investigates a method to obtain sensor location and orientation invariance for classification of wrist-mounted accelerometry via a technique known as domain adaption. Methods: Data was gathered from 16 participants who wore accelerometers on both wrists. Physical activity classification models were created using data from each wrist and then used to predict activities when using data from the opposing wrist. Using subspace alignment domain adaption, this procedure was then repeated to align the training and testing data before the classification stage. Results: Prediction of activity when using data where the wearer’s wrist was incorrectly specified resulted in a significant (p = .01) decrease in performance of 12%. When using domain adaption this drop in performance became negligible (M difference < 1%, p = .73). Conclusion: Domain adaption is a valuable method for achieving accurate physical activity classification independent of sensor orientation in wrist-worn accelerometry.

Twaites and Hillsdon are with the Sport and Health Sciences Faculty; Everson is with the Computer Sciences Faculty; University of Exeter, Exeter, United Kingdom. Langford is with Activinsights Ltd., Kimbolton, United Kingdom.

Twaites (Jt517@exeter.ac.uk) is corresponding author.
Journal for the Measurement of Physical Behaviour
Article Sections
References
  • AllenN.SudlowC.DowneyP.PeakmanT.DaneshJ.ElliottP. . . . SprosenT. (2012). UK Biobank: Current status and what it means for epidemiology. Health Policy and Technology 1(3) 123126. doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • BanosO.GalvezJ.M.DamasM.PomaresH. & RojasI. (2014). Window size impact in human activity recognition. Sensors 14(4) 64746499. PubMed ID: 24721766 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • BaoL. & IntilleS.S. (2004April). Activity recognition from user-annotated acceleration data. In International Conference on Pervasive Computing (pp. 117). Berlin, Heidelberg: Springer.

    • Search Google Scholar
    • Export Citation
  • BreimanL. (2001). Random forests. Machine Learning 45(1) 532. doi:

  • ChowdhuryA.K.TjondronegoroD.ChandranV. & TrostS.G. (2017). Ensemble methods for classification of physical activities from wrist accelerometry. Medicine & Science in Sports & Exercise 49(9) 19651973. PubMed ID: 28419025 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • CookD.FeuzK.D. & KrishnanN.C. (2013). Transfer learning for activity recognition: A survey. Knowledge and Information Systems 36(3) 537556. PubMed ID: 24039326 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • CsurkaG. (Ed.). (2017). Domain adaptation in computer vision applications (pp. 135). Springer. doi:

  • EsligerD.W.RowlandsA.V.HurstT.L.CattM.MurrayP. & EstonR.G. (2011). Validation of the GENEA Accelerometer. Medicine & Science in Sports & Exercise 43(6) 10851093. PubMed ID: 21088628 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • FaircloughS.NoonanR.RowlandsA.Van HeesV.KnowlesZ. & BoddyL. (2016). Wear compliance and activity in children wearing wrist and hip mounted accelerometers. Medicine & Science in Sports & Exercise 48(2) 245253. PubMed ID: 26375253 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • FernandoB.HabrardA.SebbanM. & TuytelaarsT. (2013). Unsupervised visual domain adaptation using subspace alignment. In Proceedings of the IEEE International Conference on Computer Vision (pp. 29602967). doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • GjoreskiM.GjoreskiH.LuštrekM. & GamsM. (2016). How accurately can your wrist device recognize daily activities and detect falls?. Sensors 16(6) 800. doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • IncelO.D. (2015). Analysis of movement, orientation and rotation-based sensing for phone placement recognition. Sensors 15(10) 2547425506. PubMed ID: 26445046 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • LeverJ.KrzywinskiM. & AltmanN. (2017). Points of Significance: Principal component analysis. Nature Methods 14(7) 641642. doi:

  • MontoyeA.H.PivarnikJ.M.MuddL.M.BiswasS. & PfeifferK.A. (2016). Wrist-independent energy expenditure prediction models from raw accelerometer data. Physiological Measurement 37(10) 17701784. PubMed ID: 27653642 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • PaveyT.G.GilsonN.D.GomersallS.R.ClarkB. & TrostS.G. (2017). Field evaluation of a random forest activity classifier for wrist-worn accelerometer data. Journal of Science and Medicine in Sport 20(1) 7580. PubMed ID: 27372275 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • PaveyT.G.GomersallS.R.ClarkB.K. & BrownW.J. (2016). The validity of the GENEActiv wrist-worn accelerometer for measuring adult sedentary time in free living. Journal of Science and Medicine in Sport 19(5) 395399. PubMed ID: 25956687 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • PedregosaF.VaroquauxG.GramfortA.MichelV.ThirionB.GriselO. . . . VanderplasJ. (2011). Scikit-learn: Machine learning in Python. Journal of Machine Learning Research 12(2011) 28252830.

    • Search Google Scholar
    • Export Citation
  • PrinceS.A.ReedJ.L.MarkA.E.BlanchardC.M.GraceS.L. & ReidR.D. (2015). A comparison of accelerometer cut-points among individuals with coronary artery disease. PLoS One 10(9) e0137759. PubMed ID: 26361345 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • StaudenmayerJ.PoberD.CrouterS.BassettD. & FreedsonP. (2009). An artificial neural network to estimate physical activity energy expenditure and identify physical activity type from an accelerometer. Journal of Applied Physiology 107(4) 13001307. PubMed ID: 19644028 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • TrostS.G.ZhengY. & WongW.K. (2014). Machine learning for activity recognition: hip versus wrist data. Physiological Measurement 35(11) 21832189. PubMed ID: 25340887 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • YangJ.NguyenM.N.SanP.P.LiX. & KrishnaswamyS. (2015July). Deep convolutional neural networks on multichannel time series for human activity recognition. In Proceedings of the 24th International Conference on Artificial Intelligence (Vol. 15 pp. 39954001).

    • Search Google Scholar
    • Export Citation
  • ZhangS.RowlandsA.V.MurrayP. & HurstT.L. (2012). Physical activity classification using the genea wrist-worn accelerometer. Medicine & Science in Sports & Exercise 44(4) 742748. PubMed ID: 21988935 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
Article Metrics
All Time Past Year Past 30 Days
Abstract Views 29 29 1
Full Text Views 1 1 1
PDF Downloads 1 1 1
Altmetric Badge
PubMed
Google Scholar