İnsan Aktivitelerinin Sınıflandırılmasında Tekrarlayan Sinir Ağı Kullanan Derin Öğrenme Tabanlı Yaklaşım

Bireyler üzerinden istenildiği anda insan aktivitelerini sınıflandırma ve tanıma sistemleri ile bilgi elde edilebilmektedir. Bu sistemler hastalıkların tespiti, fizik tedavi aşamalarının iyileştirilmesi, akıllı ev projelerinin geliştirilmesi gibi farklı alanlarda kullanılmaktadır. Bu çalışmada akıllı telefonlardaki ivmeölçer ve jiroskop duyargalarından elde edilmiş halkın kullanımına açık bir veri kümesinden alınan veriler kullanılmıştır. Literatürdeki çalışmaların çoğu yapay sinir ağı modeliyle zaman serilerinin işlenmesine dayanan daha yüksek seviyeli öznitelikleri ve bunların aralarındaki ilişkileri çözümleyememektedir. Uzun-Kısa Süreli Bellek (UKSB) modeli tekrarlayan sinir ağı olarak hem zaman serileri için ilişki elde edebilmesi hem de katmanlar halinde kullanılabilen esnek yapısı nedeniyle oldukça uygun bir derin öğrenme yaklaşımıdır. Bu altyapıyı içeren derin öğrenme tabanlı yaklaşım çalışmamızdaki deneylerde çeşitli insan aktivitelerinin sınıflandırılmasında kullanılmıştır. Deneylerde farklı girdi parametreleri, katman ve ağ birimleri ilgili ağ modellerine verilerek sınıflandırma başarımı doğruluk oranı ölçülmüştür. Sonuçta yaklaşık %86 ilâ %93 arasında sınıflandırma başarımı elde edilerek altı farklı sınıfın yüksek doğrulukta sınıflandırıldığı gösterilmiştir. Çalışmada buna dair tartışma ve elde edilen bilimsel bulgulara da yer verilmektedir.

Deep Learning-based Approach using Recurrent Neural Network for Classification of the Human Activities

Information can be obtained through classification and recognition systems of human activities at any time. These systems are used in different areas such as disease detection, improvement of physical therapy stages, development of smart home projects, and etc. In this study, data taken from a public data set obtained from accelerometer and gyroscope sensors in smart phones were used. Most of the studies in the literature cannot analyze higher level attributes and their relationships based on time series processing with artificial neural network model. The Long-Short Term Memory (LSTM) model is a very suitable deep learning approach due to its ability to obtain relationships for time series as a recurrent neural network and to be flexible in its layers. The deep learning-based approach that includes this infrastructure has been used in the classification of various human activities in our experiments. In the experiments, different input parameters, layer and network units were given to related network models and classification performance accuracy rate was measured. As a result, a classification performance of approximately 86% to 93% was obtained, showing that six different classes were classified with high accuracy. Discussion and scientific findings are also included in the study.

___

  • Y. Chen, C. Shen, Performance analysis of smartphone-sensor behavior for human activity recognition, IEEE Access 5 (2017) 3095–3110.
  • A. Campbell, T. Choudhury, From smart to cognitive phones, IEEE Pervasive Comput. 11 (3) (2012) 7–11.
  • B.P. Clarkson. Life patterns: Structure from wearable sensors (Ph.D. thesis), Massachusetts Institute of Technology, 2002.
  • Zhang, Z., L. Ji, Huang, Z. ve Wu, J., (2011). “Multi-Model Adaptation for Thigh Movement Estimation Using Accelerometers”, IET Signal Procesing, 5:709-716.
  • Muscillo, R., Schmid, M., Conforto, S. ve D’Alessio T., (2010). “An Adaptive Kalman-Based Bayes Estimation Technique to Classify Locomotor Activities in Young and Elderly Adults Through Accelerometers”, Medical Engineering & Physics, 32:849-859.
  • Yanga, M., Zhenga, H., Wanga, H., McCleanb, S. ve Newellc, D., (2012). “iGAIT: An interactive accelerometer based gait analysis system”, Computer Methods and Programs in Biomedicine, 108:715-723.
  • Chung, P.Y.M. ve Ng, G.Y.F., (2012). “Comparison Between an Accelerometer and a Three-Dimensional Motion Analysis System for The Detection of Movement”, Physiotherapy, 98:256-259.
  • Mathie vd., (2004). “Classification of basic Daily movements using a triaxial accelerometer”, Medical & Biological Engineering & Computing, 42:679-687.
  • A. Avci, S. Bosch, M. Marin-Perianu, R. Marin-Perianu, P. Havinga, Activity recognition using inertial sensing for healthcare, wellbeing and sports applications: A survey, in: 2010 23rd International Conference on Architecture of Computing Systems (ARCS), pp. 1-10. VDE (2010)
  • W. Lin, M.-T. Sun, R. Poovandran, Z. Zhang, Human activity recognition for video surveillance, in: 2008 IEEE International Symposium on Circuits and Systems, 2737-2740
  • Yang, J. Y., Wang, J. S. and Chen, Y. P. (2008). Using acceleration measurements for activity recognition: An effective learning algorithm for constructing neural classifiers. Pattern recognition letters, 29(16), 2213-2220.
  • Ravi, N., Dandekar, N., Mysore, P. and Littman, M.L. (2005). Activity recognition from accelerometer data. AAAI, 5, 1541-1546.
  • De la Vega, L.G.M., Raghuraman, S., Balasubramanian, A., and Prabhakaran, B. (2013). Exploring unconstrained mobile sensor based human activity recognition. 3rd International Workshop on Mobile Sensing, 8-11 April 2013, Philadelphia USA.
  • Sağbaş E.A, Ballı S. (2016). Akıllı Telefon Algılayıcıları ve Makine Öğrenmesi Kullanılarak Ulaşım Türü Tespiti. Pamukkale Üniversitesi Mühendislik Bilimleri Dergisi, Cilt 22, Sayı 5, Oca 2016, Sayfalar 376- 383
  • Jin Wang, Ronghua Chen, Xiangping Sun. Generative models for automatic recognition of human daily activities from a single triaxial accelerometer[C]. WCCI 2012 IEEE World Congress on Computational Intelligence June, 2012: 10-15
  • L. Bao, S. Intille, Activity recognition from user-annotated acceleration data, Pervasive Computing (2004). pp 1-17
  • J. Aggarwal, M.S. Ryoo, Human activity analysis: A review, ACM Comput. Surv. 43 (3) (2011) 1–16.
  • J.L. Reyes-Ortiz, L. Oneto, A. Sama, X. Parra, D. Anguita, “Transition-Aware Human Activity Recognition Using Smartphones”, Neurocomputing, Volume 171 Issue C, January 2016 Pages 754-767
  • Tural K., Akdoğan E., "Akıllı Telefonların Algılayıcılarının Verilerini Kullanarak Yapay Sinir Ağları ile İnsan Hareketlerinin Sınıflandırılması", Otomatik Kontrol Ulusal Toplantısı TOK 2017, İSTANBUL, TÜRKIYE, 21-23 Eylül 2017, cilt.1, no.1, ss.479-483
  • J.L. Reyes-Ortiz, L. Oneto, A. Ghio, A. Sama, D. Anguita, X. Parra, “Human Activity Recognition on Smartphones with Awareness of Basic Activities and Postural Transitions”, Artificial Neural Networks and Machine Learning, ICANN 2014. Lect. Notes Comp. Sci. 8681, pp. 177–184, Springer, 2014.
  • X. Su, H. Tong, and P. Ji,” Activity Recognition with Smartphone Sensors”, Tsinghua Science and Technology, International Journal of Information Science, vol. 19, no. 3, pp. 235-249, 2014.
  • Sağbaş E.A, Ballı S. (2015) Akıllı Telefon Sensörlerinin Kullanımı ve Ham Sensör Verilerine Erişim. Akademik Bilişim, Eskişehir, Türkiye, 4-6 Şubat 2015, pp. 158-164
  • Gunduz, H. ve Cataltepe, Z. (2015). Borsa Istanbul (BIST) daily prediction using financial news and balanced feature selection, Expert Systems with Applications, 42(22), 9001–9011.
  • Shingo Murata, Jun Namikawa, Hiroaki Arie, Shigeki Sugano, and Jun Tani, “Learning to Reproduce Fluctuating Time Series by Inferring Their Time-Dependent Stochastic Properties: Application in Robot Learning via Tutoring,” IEEE Transactions on Autonomous Mental Development (2013 JIF: 1.348), Vol. 5, Issue 4, pp. 298–310, 2013.
  • Cevher Özden, Çiğdem Acı. Analysis of injury traffic accidents with machine learning methods: Adana case. Pamukkale Univ Muh Bilim Derg. 2018; 24(2): 266-275
  • Mikolov, Tomas, Karafiat, Martin, Burget, Lukas, Cernocky, Jan, Khudanpur, Sanjeev (2010): "Recurrent neural network based language model", In INTERSPEECH-2010, 1045-1048.
  • W. Feng, N. Guan, Y.Li, X.Zhang, Z.Luo, Audio visual speech recognition with multimodal recurrent neural networks (2017), pp. 681-688
  • Hochreiter,Sepp, and Jürgen Schmidhuber. Long short-term memory, Neural computation. 9(8) (1997)1735–1780
  • K. Greff, R. Srivastava, J. Koutnik, B. Steunebrink, J. Schmidhuber. LSTM: A Search Space Odyssey. IEEE Transactions on Neural Networks and Learning Systems, Vol 28, Issue 10, 2017. Pp 2222-2232
  • UCI veri kümesi websitesi. “Human Activity Recognition Using Smartphones Data Set”.https://archive.ics.uci.edu/ml/datasets/human+activity+recognition+using+smartphones. (03.09.2019)
  • Python programlama dili websitesi. https://www.python.org/. (03.09.2019).
  • Anaconda Veri Bilimi Platformu websitesi. https://www.anaconda.com/. (03.09.2019).
  • PyCharm IDE websitesi. https://www.jetbrains.com/pycharm/. (03.09.2019).
  • Google Tensorflow API websitesi. https://www.tensorflow.org/. (03.09.2019).
  • Keras API websitesi. https://keras.io/. (03.09.2019)