论文部分内容阅读
为提高基于单一特征检测算法的准确率和可靠性,提出基于多个特征的驾驶疲劳融合检测算法.从直接反映驾驶员疲劳的2个面部特征和间接反映疲劳的1个车辆行为特征2个方面对驾驶疲劳进行综合检测.该算法运用TS模糊神经网络来识别驾驶疲劳,采用减法聚类对网络进行结构辨识,确定模糊规则的条数及相关参数的初始值,并改进了粒子群优化算法对网络进行训练.仿真和实车实验表明,该算法不仅能有效改善TS模糊神经网络的收敛速度和识别精度,而且能提高驾驶疲劳的检测正确率.
In order to improve the accuracy and reliability of single-feature-based detection algorithm, a fusion detection algorithm based on multiple features is proposed, which is based on two face features that directly reflect the driver’s fatigue and one vehicle’s behavior feature that directly reflects the fatigue This algorithm uses TS fuzzy neural network to identify driving fatigue, and uses subtractive clustering to identify the structure of the network, determines the number of fuzzy rules and the initial values of related parameters, and improves the performance of particle swarm optimization Network.The simulation and real vehicle experiments show that this algorithm can not only effectively improve the convergence speed and recognition accuracy of TS fuzzy neural network, but also improve the detection accuracy of driving fatigue.