Custom hardware architectures for posture analysis Student : M . P . T . Juvonen Course : ISE 4 Project Supervisor : Professor Wayne Luk

In this report I describe the design and implementation of an fpga-based architecture for video processing and specifically posture analysis. The final architecture performs posture analysis at a rate of 167 frames per second (projected) using frame size of 320×240 pixels, or 41 frames per second for dvd quality (720× 576 pixels) frames. This represents a 21-fold speedup over a software version running on a 3 GHz Pentium 4. This project has the following achievements: 1. Specified a set of requirements for a hardware-based posture analysis system. 2. Designed a workflow for a posture analysis system that fulfils these requirements, and designed an extensible C++ software framework for constructing the posture analysis workflow. 3. Designed a parameterised, extensible hardware architecture for image processing to allow for easy design, testing, implementation and improvement of the posture analysis system. 4. Used this architecture to design and implement a fully pipelined, parameterisable posture analysis architecture. The architecture can use multiple posture description algorithms to obtain different, complementary descriptions from a single source image.