人工神经网络对电子鼻性能的影响

被引:10
作者
秦树基
徐春花
王占山
机构
[1] 同济大学物理系
关键词
反向传输网络; 学习矢量量化网络; 概率神经网络; 模式识别; 电子鼻;
D O I
暂无
中图分类号
TP24 [机器人技术];
学科分类号
080202 ; 1405 ;
摘要
电子鼻原型由4个气体传感器组成的阵列和人工神经网络识别软件组成,可识别不同品牌的白酒.以它为例,研究了3种人工神经网络,即反向传输网络(BPN)、学习矢量量化网络(LVQ)和概率神经网络(PNN)对电子鼻性能的影响.结果表明,在需要精细识别时,虽然传感器阵列对白酒的响应谱的差别是电子鼻识别的基础,但是人工神经网络结构和算法包括相关训练参数的选择对决定电子鼻的性能也有重要的作用.比较而言,学习矢量量化网络在分类能力和训练成本方面更胜一筹,而概率神经网络则在计算负载和易用性方面更好一些.
引用
收藏
页码:804 / 808
页数:5
相关论文
共 9 条
[1]  
Chemical imaging: Trends in practical multi-parameters sensor Systems. Weimar U,Gopel W. Sensors and Actuators . 1998
[2]  
An algorithm for least squares estimation of nonlinear parameters. Marquardt D W. Journal of the Society for Industrial and Applied Mathematics . 1963
[3]  
A method for solution of certain problems in least squares. Levenberg K. Quarterly of Applied Mathematics . 1944
[4]  
The self-organizing map. Kohonen T. Proceedings of Tricomm . 1990
[5]  
Chemical imaging: Concepts and visions for electronic and bioelectronic noses. Weimar U,Gopel W. Sensors and Actuators . 1998
[6]  
A brief history of electronic noses. Gardner J W,Bartlett P N. Sensors and Actuators . 1994
[7]  
Probabilistic neural-network structure determination for pattern classification. Mao K Z,Tan K C,Ser W. IEEE Transactions on Neural Networks . 2000
[8]  
Learning vector quantization for the probabilistic neural network. Burrascano P. IEEE Transactions on Neural Networks . 1991
[9]  
Improved version of learning vector quantization. Kohonen T. Int Joint Conf on Neural Networks . 1990