A single-chip 600-fps real-time action recognition system employing a hardware friendly algorithm

A real-time action recognition system has been developed. In order to achieve an efficient and compact system implementation, a hardware friendly algorithm has been explored and employed in building the system. The core functions of the system, namely the whole action recognition algorithm including the motion detection, motion feature vector generation, nearest neighbor search, are entirely realized on a single FPGA chip. Operating at 63 MHz, the system can process QVGA (320 × 240) videos at a speed of 600-fps. The robustness of the system is demonstrated by real environment experiments on gesture recognition.

[1]  Masatoshi Ishikawa,et al.  Real-time visual measurements using high-speed vision , 2004, SPIE Optics East.

[2]  T. Shibata,et al.  A real-time image-feature-extraction and vector-generation VLSI employing arrayed-shift-register architecture , 2005, Proceedings of the 31st European Solid-State Circuits Conference, 2005. ESSCIRC 2005..

[3]  Thomas Serre,et al.  A Biologically Inspired System for Action Recognition , 2007, 2007 IEEE 11th International Conference on Computer Vision.

[4]  Tadashi Shibata,et al.  A real-time image recognition system using a global directional-edge-feature extraction VLSI processor , 2009, 2009 Proceedings of ESSCIRC.

[5]  Narayanan Vijaykrishnan,et al.  Accelerating neuromorphic vision algorithms for recognition , 2012, DAC Design Automation Conference 2012.

[6]  Zhu Li,et al.  Real-time human action recognition by luminance field trajectory analysis , 2008, ACM Multimedia.

[7]  Ronen Basri,et al.  Actions as Space-Time Shapes , 2007, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[8]  Masatoshi Ishikawa,et al.  Dynamic regrasping using a high-speed multifingered hand and a high-speed vision system , 2006, Proceedings 2006 IEEE International Conference on Robotics and Automation, 2006. ICRA 2006..

[9]  Tadashi Shibata,et al.  An image representation algorithm compatible with neural-associative-processor-based hardware recognition systems , 2003, IEEE Trans. Neural Networks.

[10]  Tadashi Shibata,et al.  A hardware friendly algorithm for action recognition using spatio-temporal motion-field patches , 2013, Neurocomputing.

[11]  Koichi Hashimoto,et al.  Two-dimensional tracking of a motile micro-organism allowing high-resolution observation with various imaging techniques , 2005 .

[12]  Hongying Meng,et al.  Real-time human action recognition on an embedded, reconfigurable video processing architecture , 2008, Journal of Real-Time Image Processing.

[13]  Tadashi Shibata,et al.  A Digital-Pixel-Sensor-Based Global Feature Extraction Processor for Real-Time Object Recognition , 2009 .