自组织RBF 神经网络对驾驶员主动安全性因素的辨识

利用自组织RBF神经网络对驾驶员主动安全性因素进行辨识。对网络进行训练时，首先由改进FCM算法根据输入样本内部关系确定RBF参数并通过一个聚类合理性聚类控制聚类个数，然后根据网络映射性能对RBF参数细调并对输出权值学习。利用本文给出的网络成功地对驾驶员的熟练程度进行了识别。

Identification of the Driver 's Active Safety Factors Using Self organizing RBF Neural Network

A self-organizing RBF neural newtork (SORBFNN) is presented to identify the drivers degrees of proficiency. A novel fuzzy c-means (FCM) algorithm is used to estimate the parameters of RBFs. The number of RBFs is determined by the clustering rationality function. A supervised learning scheme based on the class-conditional variance and the mapping error is proposed to tune the parameters of RBFs fine and the output weight values. The proposed learning algorithm estimates the clusters of the feature vectors rationally as well as increases the mapping ability of the SORBFNN. The SORBFNN is trained and tested on the experimental data of the drivers， proficiency with satisfactory results.