基于DSP的驾驶员疲劳实时预警系统设计

Design of Driver Fatigue Real-time Early Warning System Based on DSP

为设计并且实现了一个基于DSP的嵌入式车载疲劳检测系统。构建了由CCD,TMS320DM6437,红外光源组成的硬件平台。结合PERCLOS模型来进行驾驶员疲劳程度的判断。针对AdaBoost算法中不适合TMS320C6437处理器特点的大量浮点运算,在保持AdaBoost算法精度的基础上,提出了一套浮点数转定点数的代码优化方案。经过大量实验,对比其相关领域研究成果,该系统具有在黑夜或光线变化强烈的情况下依然保持检测精度,可靠性高、实时性好的特点,驾驶员疲劳检测系统能够在实际的交通运输中应用。

In this paper,a DSP-based,embedded vehicle fatigue detection system is designed and realized.A hardware platform is constructed which is composed by CCD,TMS3206437 and infrared light,combined with PERCLOS model to judge the degree of driver fatigue.For the Adaboost algorithm has a large number of floating-point operations,which is not suitable for the characteristics of TMS320C6437,a floating-point to fixed-point plan is advanced without losing any accuracy that the original one should have in this paper.The experimental results show that driver fatigue detection system can detect drivers’ eye movement and blink information in real-time.This system can still work well in darkness or light-strong condition,and has high reliability,real-time.It is available in traffic.