Research On Hardware Architecture Of Hand Tracking System And Its Implementation With FPGA

Since the beginning of the21stCentury, multimedia technology has developed rapidlyand been widely used. Instead of depending on traditional equipment, a new generation ofhuman-computer interaction(HCI) techniques are being rapidly developed by using a varietyof human sensory channels (such as voice input, gesture input, movement input, posture input,etc.) to interact with computer in a parallel, high-speed, and non-precise way. However, undera complex background, the application and industrialization for real-time, non-contact HCIsystem is still at a relatively early stage in current China.In this article, a set of formed gesture recognition tracking system is studied. Toovercome the problem brought by large data computation and high data throughput to thealgorithm, this article states how to use pure hardware scheme to implement the wholealgorithm system through making full use of the characteristic of FPGA(Field-ProgrammableGate Array) parallel computation. Moreover, this article pays attention to how to solve theproblems of system hardware architecture design, the improvement of algorithm for hardwareimplementation, storage cache design bottlenecks of high data throughput system, algorithmmodule hardware design, system validation assessment program and so on. This designimproves the system operator parallelism and utilization of hardware resources, consequentlyimproves the system detecting speed and system efficiency, so as to realize prototypeverification for making this non-contact HCI system into chips.This design uses Verilog hardware description language for the whole system description.An experimental platform which consists of Terasic company’s DE4-230and DE2-70development board is applied to the whole system’s simulation, testing, verification, analysisand evaluation. After testing and comparative assessment, the result shows that this system isable to detact and track a video of640×480resolution and of30fps frame rate in real-time,achieve the real-time gesture recognition tracking with both high-speed and high-performance.Basically it has achieved almost all of the performances of algorithm system. This also showsthat interaction experience has an advantage over the verification on computer.