人体关节运动特征追踪平台的设计与实现

Design and implementation of feature tracking platform for human joint motion

人体运动关键特征的实时追踪和在线监测是模式识别领域的一个重要课题,面临跟踪误差大,跟踪滞后等问题。提出一种新的人体关节运动特征视觉追踪平台设计方法,首先得到人体关节运动特征提取和追踪的动力学模型,进行视觉跟踪系统的总体设计和功能指标分析描述,对平台的设计分为硬件设计和软件开发两部分,采用ADSP-BF537作为核心处理芯片,对系统的复位电路设计、串口电路设计、外部存储器设计和A/D接口以及电源电路进行集成设计描述,软件开发建立在CCS 2.20平台基础上,通过程序加载、初始化和中断优化设计,实现软件开发。试验结果表明,该平台进行体的关节的运动特征追踪和信息采集的准确度高,系统运行的实时性好,可靠性和稳定性较高。

Real-time tracking and online monitoring of human movement key characteristics are important subjects in the field of pattern recognition, but now they are confronted with big tracking error, tracking lag and so on. Because of these, a new design method of human body joint movement characteristics of visual tracking platform is put forward. A dynamics model of hu- man body joint motion feature extraction and tracking should be got at first for overall design and function index analysis of visu- al tracking system. The design of platform divided into two parts, hardware design and software development. The ADSP - BF537 is taken as the core processing chip to perform the system integrated design description of reset circuit design, serial in- terface circuit design and A/D interface design, external memory design and power supply circuit. The software development is based on CCS 2.20 platform. It is realized by means of program load, initialization and interrupt optimization design. The test re- sults show that the platform has the characteristics of high-accuracy movement feature tracking and information acquisition, high real-time operation performance, high reliability and stability.