Comparison of Neural Network Algorithms to Determine the Range of Motion Using Skeleton Models (original) (raw)

Abstract

We have the main characteristics of moving as walking, running, sitting, dancing, eating, drinking and other activities. The range-of-motion, ROM, is the maximum number of movements that can occur in the sagittal, frontal, and transverse type and consists of flexion, extension, abduction, adduction, hyperextension and others. Decreased ROM can be the result of injury and the aging process and can lead to undesirable motion patterns. Feature extraction in motion analysis involve calculating a number of characteristic values independent of the size to produce the appropriate motion identification using the best method or algorithm during processing. Based on the data types obtained from the process of extracting properties collected using moment invariance, the results of identification based on ROM standard gave an accuracy of 98% using Back propagation Neural Network, 94% using Radial Basis Function Neural Network and 36% using K Means Clustering.

Key takeaways

sparkles

AI

  1. Back Propagation Neural Network achieved 98% accuracy for Range of Motion identification.
  2. The study focuses on analyzing human motion using skeleton models and neural networks.
  3. Feature extraction utilized moment invariants for robust motion recognition.
  4. K Means Clustering performed poorly with only 36% accuracy in motion identification.
  5. Data was captured using high-resolution cameras at 30 frames per second.

Loading...

Loading Preview

Sorry, preview is currently unavailable. You can download the paper by clicking the button above.

References (21)

  1. Calculate moment M1=n20+n02; M2=(n20-n02)^2+4*n11^2; M3=(n30-3*n12)^2+(3*n21-n03)^2;
  2. M4=(n30+n12)^2+(n21+n03)^2;
  3. M5=(n30-3*n21)*(n30+n12)*[(n30+n12)^
  4. *[3*(n30+n12)^2-(n21+n03)^2];
  5. +4*n11*(n30+n12)*(n21+n 03 M7=(3*n21-
  6. *(n30+n12)*[(n30+n12)^2-3*(n21+n03)^2]- (n30+3*n12)*(n21+n03)*[3*(n30+n12)^2-
  7. Show Result moment invariant M1-M7 M= [M1 M2 M3 M4 M5 M6 M7]
  8. If vid Width = Width and if Height = .Height;
  9. mov = truct('cdata',zeros(vidHeight,vidWidt h,3,'uint8'),.'color map',[]);
  10. write(v(k).c (['Frame ',num2str(k),'.jpg']))
  11. Arno HA, Arvid QL, "Visualization of Shoulder Range of Motion for Clinical Diagnostics and Device Development", IEEE International Conference on Rehabilitation Robotic (ICORR), 2015, pp. 816-821, DOI: 10. 1109 / ICORR 2015. 7218303.
  12. A.Vaishnavdevi, S.Suganraj, L. Mareeswaran, S. Subaraja, K.Srinivas, 2017, Analysis of Brodatz Image Using GLCM and Hu's Moments, SSRG International Journal of Electronics and Communication Engineering -(ICCREST'17) -Special Issue - March , PP.54-58, ISSN: 2348 -8549.
  13. Huang, M.-C., Lee, S.-H., Yeh, S.-C., Chan, R.-C., Rizzo, A., Xu, W., Shan-Hui, L. (2014). Intelligent Frozen Shoulder Rehabilitation. IEEE Intelligent Systems, 29(3), 22-28. doi:10.1109/mis.2014.35.
  14. Pasquale Daponte,2013, Validation of a Home Rehabilitation System for Range of Motion Measurements of Limb Functions, 978-1-4673- 5197-3/13/$31.00 ©2013 IEEE, IEEE Fellow Member, Luca De Vito, IEEE Senior Member, Carmine Sementa Department of Engineering, University of Sannio, Benevento, Italy {daponte, devito, csementa}@unisannio.it
  15. Fauziah, Eri Prasetyo Wibowo, Hustinawaty, Sarifuddin Madenda, "Identification of hand motion using background subtraction method and extraction of image binary with back propagation neural network on skeleton model", 2nd International Conference on Computing and Applied Informatics 2017, IOP Conference Series: Journal of Physic: Conference .Series 978(2018) 012020 1-14, DOI:10.1088/1742- 6596/978/012020.
  16. Sara Amendola, Luigi Biachi, Gaetano Marocco,2015, Movement Detection Of Human Body Segments,ISSN: 1045-9243, DOI: 10.1109/MAP.2015.2437274, PP: 23 -37.
  17. Rafael C Gonzalez, Richard E Woods, 2002, "Digital Image Processing", New Jersey Prentice Hall, ISBN-13: 978-0201180756.
  18. Ferda Ofli, Rizwan Chaudhry, Gregorij Kurillo, Rene Vidal And Ruzena Bajcsy, 2012,"Sequence Of The Most Informative Joints (SMIJ): A New Representation For Human Skeletal Action Recognition", Computer Vision and Pattern Recognition Workshops (CVPRW), 2012 DOI: 10.1109/CVPRW.2012.6239231, PP: 8-12.
  19. G.A. Papakostas, D.E. Koulouriotis, V.D. Tourassis (2012). Feature Extraction Based on Wavelet Moments and Moment Invariants in Machine Vision Systems, Human-Centric Machine Vision, ISBN:978-953-51-0563-3.
  20. Tavipia Rumambi, Hustinawaty, Sarifuddin Madenda, Eri Prasetyo Wibowo, 2017, Measurement Straight Leg Raise for Low Back Pain Based Grayscale Depth",TELKOMNIKA, Vol.15, No.1, March 2017, pp. 471~477 ISSN: 1693-6930.
  21. Tae Soo Kim, Austin Reiter, 2017, "Interpretable 3D Human Action Analysis with Temporal Convolutional Networks", Published in: Computer Vision and Pattern Recognition Workshops (CVPRW), 2017 IEEE Conference on Honolulu, HI, USA, Date of Conference: 21-26 July 2017, IEEE 24 August 2017, ISBN Information: Electronic ISSN: 2160-7516, 1623-1631,DOI: 10.1109/CVPRW.2017.207.