Fiziksel Hareketlerden Aktivite Belirleme Yöntemlerinin Karşılaştırılması

Bu çalışmada, atalet sensör (IMU) verilerinden kişilerin temel fiziksel hareketlerini (aşağı inme, yukarı çıkma, oturma, durma, yürüme, koşma) otomatik algılayabilen yöntemler incelenmiştir. Klasik (YSA, DVM, kNN) ve güncel yaklaşımların (Evrişimsel Sinir Ağları-ESA) IMU verilerini aktivite sınıflarına eşleme performansları karşılaştırılmıştır. Bunun için üç aşamalı bir çalışma gerçekleştirilmiştir: 1) veri elde etme; 2) eğitim/test kümelerini oluşturma; 3) ağ mimarilerinin inşası ve sınıflandırma. Veri elde etme aşamasında, üzerine ivmeölçer sensörü yerleştirilen 10 farklı kişiden 6 farklı fiziksel hareketi tekrarlı bir şekilde gerçekleştirmesi sağlanmış ve hareket esnasında ivmeölçer verileri kaydedilmiştir. İkinci aşamada, kaydedilen uzun süreli ivmeölçer verileri kısa süreli pencereler şeklinde paketlere bölünmüştür. Tek boyutlu ivme bilgilerini içeren her bir paket verisinden öznitelikler çıkarılarak klasik yaklaşımların eğitim kümesi inşa edilmiştir. Güncel yaklaşımın eğitim kümesi için tek boyutlu sinyallerin iki boyutlu görüntü matrisine dönüşümü gerçekleştirilmiştir. Üçüncü aşamada, YSA, DVM, k-NN ve ESA mimarileri inşa edilerek sınıflandırma işlemi gerçekleştirilmiştir. Yapılan deneysel çalışmalar sonucunda IMU-aktivite eşleştirme doğruluğunun YSA yöntemiyle %99, ESA yöntemiyle %95 olduğu görülmüştür.

Comparison of methods for determining activity from physical movements

In this study, the methods which can detect the basic physical movements of a person (downward, upward, sitting, stop, walking,running) from inertial sensor (IMU) data are evaluated. The performances of classical (ANN, SVM, k-NN) and current approaches(Convolutional Neural Networks-ESA) to map IMU data to activity classes were compared. A three-stage study was carried outfor this aim: 1) data acquisition; 2) creating training/test sets; 3) construction and classification of network architectures. At thestage of data acquisition, to obtain 6 different physical movements from 10 different people, the accelerometer sensor is placed onthe persons. Repetitive movements of persons were recorded. At the second stage, the recorded long-term accelerometer data isdivided into packages in the form of short-term windows. The training set of classical approaches was constructed by featuresextracting from each packet data containing one-dimensional acceleration information. The transformation of one-dimensionalsignals to a two-dimensional image matrix for the training set of the deep learning-based approaches was performed. In the thirdstage, ANN, SVM, k-NN and CNN architectures were constructed, and classification process was carried out. As a result of theexperimental studies, it was found that the accuracy of IMU-activity mapping was 99% with the ANN method and 95% with theCNN method.

PDF

___

[1] Lane N. D., Miluzzo E., Lu H., Peebles D., Choudhury T. and Campbell A. T., “A survey of mobile phone sensing”, IEEE Commun. Mag., 48(9): 140–150, (2010).
[2] Bourke A. K., Ihlen E. A. F., Van De Ven P., Nelson J. and Helbostad J. L., “Video analysis validation of a realtime physical activity detection algorithm based on a single waist mounted tri-axial accelerometer sensor”, Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), 4881–4884,(2016).
[3] Han Y. C., Wong K. I. and Murray I., “Automatic synchronization of markerless video and wearable sensors for walking assessment”, IEEE Sens. J., 19(17): 7583–7590, (2019).
[4] Prasertsung P. and Horanont T., “A classification of accelerometer data to differentiate pedestrian state”, 20th International Computer Science and Engineering Conference: Smart Ubiquitos Computing and Knowledge (ICSEC),(2017).
[5] Su X., Tong H. and Ji P., “Activity recognition with smartphone sensors”, Tsinghua Sci. Technol., 19(3): 235–249, (2014).
[6] Silva de Lima A. L. et al., “Freezing of gait and fall detection in Parkinson’s disease using wearable sensors: a systematic review”, Journal of Neurology., 264: 1642– 1654, (2017).
[7] Pereira A., Folgado D., Nunes F., Almeida J. and Sousa I., “Using Inertial Sensors to Evaluate Exercise Correctness in Electromyography-based Home Rehabilitation Systems”, Medical Measurements and Applications, MeMeA 2019 - Symposium Proceedings, (2019).
[8] Sağbaş E. A. and Balli S., “Transportation mode detection by using smartphone sensors and machine learning”, Pamukkale Univ. J. Eng. Sci., 22(5): 376–383, (2016).
[9] Kwapisz J. R., Weiss G. M. and Moore S. A., “Activity recognition using cell phone accelerometers”, ACM SIGKDD Explor. Newsl., 12(2):. 74, (2011).
[10] Shoaib M., Bosch S., Scholten H., Havinga P. J. M. and Incel O. D., “Towards detection of bad habits by fusing smartphone and smartwatch sensors”, 2015 IEEE International Conference on Pervasive Computing and Communication Workshops, PerCom Workshops 2015, 591-596 (2015).
[11] Natarajasivan D. and Govindarajan M., “Filter Based Sensor Fusion for Activity Recognition using Smartphone”, Int. J. Comput. Sci. Telecommun. J. Homepage, 7(5): (2016).
[12] Dadashi F. et al., “A hidden Markov model of the breaststroke swimming temporal phases using wearable inertial measurement units”, 2013 IEEE International Conference on Body Sensor Networks (BSN 2013), (2013).
[13] Mortazavi B. J., Pourhomayoun M., Alsheikh G., Alshurafa N., Lee S. I. and Sarrafzadeh M., “Determining the single best axis for exercise repetition recognition and counting on smartwatches”, Proceedings - 11th International Conference on Wearable and Implantable Body Sensor Networks, (BSN 2014), 33–38, (2014).
[14] Guiry J. J., Van de Ven P. and Nelson J., “Multi-sensor fusion for enhanced contextual awareness of everyday activities with ubiquitous devices”, Sensors (Switzerland), 14(3): 5687–5701, (2014).
[15] Weiss G. M., Timko J. L., Gallagher C. M., Yoneda K. and Schreiber A. J., “Smartwatch-based activity recognition: A machine learning approach”, 3rd IEEE EMBS International Conference on Biomedical and Health Informatics (BHI 2016), 426–429, (2016).
[16] “MTw Avinda Software”: https://www.xsens.com/mtsoftware-suite-mtw-awinda/. [Accessed: 10-Jul-2019].
[17] Xin X., Wang C., Ying X. and Wang B., “Deep community detection in topologically incomplete networks”, Phys. A Stat. Mech. its Appl., 469: 342-352, (2017).
[18] Buttrey S. E. and Karo C., “Using k-nearest-neighbor classification in the leaves of a tree”, Comput. Stat. Data Anal., 40(1):27-37, (2002).
[19] Altın M., Kazan A. and Karadag H. B., “Rotational Surfaces Generated by Planar Curves in E3 with Density”, Int. J. Anal. Appl., 17(3): 311–328, (2019).
[20] Hajibandeh N., Faghihi F., Ranjbar H. and Kazari H., “Classifications of disturbances using wavelet transform and support vector machine”, Turkish J. Electr. Eng. Comput. Sci., 25(2): 832-843, (2017).
[21] Shah S. M. S., Batool S., Khan I., Ashraf M. U., Abbas S. H. and Hussain S. A., “Feature extraction through parallel Probabilistic Principal Component Analysis for heart disease diagnosis”, Phys. A Stat. Mech. its Appl.,482: 796-807, (2017).
[22] Moraes R., Valiati J. F. and Gavião Neto W. P., “Document-level sentiment classification: An empirical comparison between SVM and ANN”, Expert Syst. Appl., 40(2): 621–633, (2013).
[23] Kingma D. P. and Ba J. L., “Adam: A method for stochastic gradient descent”, ICLR Int. Conf. Learn. Represent., (2015).