摘要
为了使交通管理系统能进行可靠的机动车分类,研究了轿车、轻型越野车和货车3种机动车目标的声信号,提出了一种采用子波分解后不同尺度上声信号能量作为特征向量的特征提取算法,并设计了kNN(k近邻)分类器和改进BP神经网络分类器用于目标分类。目标识别和分类试验结果表明:所提出的特征提取算法能够很好地体现不同类型目标之间的差异,提取的特征向量稳健;设计的改进BP神经网络分类器的分类精度可达92.6%,且分类效果优于kNN分类器。
In order to get a reliable vehicle classification for the traffic management system, target acoustic signals of three kinds of vehicles, ie car, light cross-country vehicle, and truck were studied. A feature extraction algorithm was proposed which took the time-domain energy of the acoustic signals in different scales after wavelet decomposition as the feature vectors, k-nearest neighbor classifier and improved BP neural network classifier for vehicle target classification were designed. The target recognition and classification experiment results show that the proposed feature extraction algorithm can distinguish different types of vehicles with satisfactory rate of correct recognition, and feature vector is robust. The classification accuracy of improved BP neural network classifier can reach 92. 6%, and classification performance is better than kNN classifier.
出处
《中国公路学报》
EI
CAS
CSCD
北大核心
2007年第3期97-102,共6页
China Journal of Highway and Transport
基金
国防科技重点实验室基金项目(51454070204HK0320)
西北工业大学科技创新基金项目(2003CR080001)
关键词
交通工程
机动车车型识别
子波尺度空间能量
BP神经网络
声信号
特征提取
traffic engineering
vehicle type recognition
energy in wavelet scale space
BP neural network
acoustic signal
feature extraction
