应用于卟啉阵列气体检测系统的识别算法  

Gas Recognition Algorithm Applied in Porphyrin Array Gas Detection System

作　　者：罗小刚[1] 谢新莘 

机构地区：[1]重庆大学生物工程学院,重庆400030

出　　处：《传感技术学报》2014年第11期1464-1468,共5页Chinese Journal of Sensors and Actuators

基　　金：国家科技支撑计划项目(2012BAI19B03);国家自然科学基金项目(81271930;81171414);教育部高校博士点基金项目(20090191110030);中央高校基本科研业务费项目(CDJXS10231179;CDJSX102300)

摘　　要：差值图谱是卟啉化学传感器(PSA)芯片识别气体的根据,但同种气体的差值图谱之间会出现颜色或显色位置的差异,存在实验数据发散的问题。结合BP神经网络和粗糙集,提出了一种新的气体种类识别算法,并将之用于PSA气体检测系统中。该方法利用粗糙集的数据约简功能找到差值图谱数据中颜色变化明显的点的数据,以这些数据为输入,进行BP神经网络的训练和识别。通过实验可以证明,相对于欧氏距离聚类结果、BP神经网络识别结果,本文提出的算法对于发散的实验数据具有更高的识别精度。The difference map is the basis of identifying gases by the PSA chips. However,there are differences be-tween each difference map of a gas,which is called the“divergent problem”. A pattern recognition algorithm based on backpropagation neural network and rough set was described,which was employed in the porphyrin chemical sen-sor array integrated system. That algorithm picked up the spots whose color changed obviously using the rough set, and set their values as input of BP network. Comparing with the result of Euclidean distance clustering and BP neu-ral network identification without removing unnecessary data as input,the result of the algorithm proposed in this ar-ticle has higher identification accuracy to the divergence experimental data.

关 键 词：卟啉传感器 气体识别 神经网络 粗糙集 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]