Hand Gestures Classification with Multi-Core DTW


1 Misr International University, Faculty of computers and information, HCI-LAB, Helwan University

2 Faculty of Computer Science, Misr International University



Classifications of several gesture types are very helpful in several applications. This paper tries to address fast classifications of hand gestures using DTW over multi-core simple processors. We presented a methodology to distribute templates over multi-cores and then allow parallel execution of the classification. The results were presented to voting algorithm in which the majority vote was used for the classification purpose. The speed of processing has increased dramatically due to using multi-core processors and DTW.


[1] Shaun K. Kane, Jacob O. Wobbrock, and Richard E. Ladner. Usable gestures for blind people: Understanding preference and performance. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, CHI ’11,
pages 413–422, New York, NY, USA, 2011. ACM.
[2] Amr Alanwar, Moustafa Alzantot, Bo-Jhang Ho, Paul Martin, and Mani Srivastava. Selecon: Scalable iot device selection and control using hand gestures. In Proceedings of the Second International Conference on Internet-of-Things Design
and Implementation, IoTDI ’17, pages 47–58,New York, NY, USA, 2017. ACM.
[3] Lei Jing, Zixue Cheng, Yinghui Zhou, Junbo Wang, and Tongjun Huang. Magic ring: A selfcontained
gesture input device on finger. In Proceedings of the 12th International Conference on Mobile and Ubiquitous Multimedia, MUM ’13, pages 39:1–39:4, New York, NY, USA, 2013.ACM.
[4] Ahmad Sufril Azlan Mohamed, Effha Binti Soter,Anand Singh, and Nur Intan Raihana Ruhaiyem. Gesture based help identification for hospital & elderlycare using dynamic time warping: A systematic study. In Proceedings of the International Conference on Video and Image Processing, ICVIP 2017, pages 94–98, New York, NY,USA, 2017. ACM.
[5] H. Sakoe and S. Chiba. Dynamic programming algorithm optimization for spoken word recognition. IEEE Transactions on Acoustics, Speech, and Signal Processing, 26(1):43–49, Feb 1978.
[6] Pat Jangyodsuk, Christopher Conly, and Vassilis Athitsos. Sign language recognition using dynamic time warping and hand shape distance based on histogram of oriented gradient features. In Proceedings of the 7th International Conference
on PErvasive Technologies Related to Assistive Environments, PETRA ’14, pages 50:1–50:6, New York, NY, USA, 2014. ACM.
[7] A. Atia, S. Takahashi, and J. Tanaka. Smart gesture sticker: Smart hand gestures profiles for daily objects interaction. In 2010 IEEE/ACIS 9th International Conference on Computer and Information Science, pages 482–487, Aug 2010.
[8] Aya Hamdy Ali, Ayman Atia, and Mostafa-Sami M. Mostafa. Recognizing driving behavior and road anomaly using smartphone sensors. Int. J. Ambient Comput. Intell., 8(3):22–37, July 2017.
[9] Krzysztof Pietroszek, Phuc Pham, and Christian Eckhardt. Cs-dtw: Real-time matching of multivariate spatial input against thousands of templates using compute shader dtw. In Proceedings of the 5th Symposium on Spatial User
Interaction, SUI ’17, pages 159–159, New York,NY, USA, 2017. ACM.
[10] Salwa O. Slim, Ayman Atia, and Mostafa-Sami M. Mostafa. An experimental comparison between seven classification algorithms for activity recognition. In Tarek Gaber, Aboul Ella Hassanien, Nashwa El-Bendary, and Nilanjan Dey, editors, The 1st International Conference on Advanced Intelligent System and Informatics (AISI2015), November 28-30, 2015, Beni Suef,Egypt, pages 37–46, Cham, 2016. Springer International Publishing.
[11] Maher Khemakhem, Abdelfettah Belghith, and Labidi Mohamed. The dtw data distribution over a grid computing architecture. International Journal of Computer Sciences and Engineering Systems (IJCSES), pages 241–247, 12 2007.
[12] Stan Salvador and Philip Chan. Toward accurate dynamic time warping in linear time and space. Intelligent Data Analysis, 11(5):561–580, 2007.
[13] Diego F. Silva, Rafael Giusti, Eamonn Keogh, and Gustavo E. Batista. Speeding up similarity search under dynamic time warping by pruning unpromising alignments. Data Min. Knowl. Discov.,32(4):988–1016, July 2018.
[14] Nguyen Cong Thuong and Duong Tuan Anh. Comparing three lower bounding methods for dtw in time series classification. In Proceedings of the Third Symposium on Information and Communication Technology, SoICT ’12, pages 200–206, New York, NY, USA, 2012. ACM.
[15] Kyeongeun Seo and Hyeonjoong Cho. Airpincher: A handheld device for recognizing delicate midair hand gestures. In Proceedings of the Adjunct Publication of the 27th Annual ACM Symposium on User Interface Software and Technology,
UIST’14 Adjunct, pages 83–84, New York, NY,USA, 2014. ACM.
[16] Florian Grützmacher, Johann-Peter Wolff, and Christian Haubelt. Exploiting thread-level parallelism in template-based gesture recognition with dynamic time warping. In Proceedings of the 2Nd International Workshop on Sensor-based Activity Recognition and Interaction, iWOAR ’15, pages 6:1–6:6, New York, NY, USA, 2015. ACM.
[17] Gautham Krishna Gudur, Karthik Subramanian Nathan, Yogesh Kumar B, Ankith A. Prabhu, Ajay Kannan, and Vineeth Vijayaraghavan. A generic multi-modal dynamic gesture recognition system using machine learning. CoRR, abs/1809.05839, 2018.

[18] Josep Maria Carmona and Joan Climent. A performance evaluation of hmm and dtw for gesture recognition. In Luis Alvarez, Marta Mejail, Luis Gomez, and Julio Jacobo, editors, Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, pages 236–243, Berlin, Heidelberg, 2012. Springer Berlin Heidelberg.
[19] Khaled Assaleh, Tamer Shanableh, and Mohammed Zourob. Low complexity classification system for glove-based arabic sign language recognition. In Tingwen Huang, Zhigang Zeng, Chuandong Li, and Chi Sing Leung, editors,
Neural Information Processing, pages 262–268,Berlin, Heidelberg, 2012. Springer Berlin Heidelberg.
[20] G. Costante, L. Porzi, O. Lanz, P. Valigi, and E. Ricci. Personalizing a smartwatch-based gesture interface with transfer learning. In 2014 22nd European Signal Processing Conference(EUSIPCO), pages 2530–2534, Sept 2014.
[21] Aya Hamdy Ali, Ayman Atia, and Mostafa Sami. A comparative study of user dependent and independent
accelerometer-based gesture recognition algorithms. In Norbert Streitz and Panos Markopoulos, editors, Distributed, Ambient, and Pervasive Interactions, pages 119–129, Cham,2014. Springer International Publishing.
[22] Sagar Hursale, Hiren Makwana, Ankit Panchal, Chintan Bhakkad, and Shubhangi Vairagar. Smartphone’s based gesture recognition in air. In Proceedings of the 2015 International Conference on Advanced Research in Computer Science Engineering & Technology (ICARCSET 2015),ICARCSET ’15, pages 49:1–49:5, New York, NY,USA, 2015. ACM.
[23] Ayman Atia, Shin Takahashi, and Jiro Tanaka. Smart gesture sticker: Smart hand gestures profiles for daily objects interaction. In Computer and Information Science, ACIS International Conference, pages 482–487. IEEE Computer Society, 2010.
[24] Seiichi Nakagawa and Hirobumi Nakanishi. Dynamic programming algorithm optimization for spoken word recognition. IEEE Transactions on Acoustics, Speech, and Signal Processing, 26(1):43–49, February 1978.
[25] Seiichi Nakagawa and Hirobumi Nakanishi. Speaker independent english consonant and japanese word recognition by a stochastic dynamic time warping method. IETE Journal of Research, 34(1):87–95, 1988.
[26] Ayman Atia, Shin Takahashi, Kazuo Misue, and Jiro Tanaka. Ubigesture: Customizing and profiling hand gestures in ubiquitous environment. In Human-Computer Interaction. Novel Interaction Methods and Techniques,13th International Conference, HCI International 2009, pages 141–151,2009.