智能嗅敏信息处理技术中神经网络方法的应用研究  

The Applied Research on Neuro-network Method in Intelligent Scent Information Processing Technology

作　　者：谭伟[1] 钱东海[1] 谢明江[2] 

机构地区：[1]上海交通大学机器人研究所,上海200030 [2]上海交通大学自动化研究所,上海200030

出　　处：《自动化仪表》1999年第12期8-10,共3页Process Automation Instrumentation

摘　　要：在建立人工嗅敏实验系统的基础上针对使用BP网络处理气敏阵列信号时出现的种种不足,引入较之优越的ML算法,并提出用非线性函数及SOM网络对气敏样本进行预处理,用分段训练的方法加速学习过程,最后得到在速度和精度上更接近实用要求的新的嗅敏信息处理合成算法.On the basis of establishing the experimental system of artificial scent sensitivity the better ML algorithm is introduced to eliminate the disadvantage on processing the gas sensitive signal by using BP network. In addition, the preprocessing of gas-sensitive samples with non-linear function and SOM network is stated. By using the method of piecewire training, the learning procedure is accelerated. Finally, the new composed algorithm for processing the scent information is obtained. The algorithm is more close to the requirement of application on speed and accuracy.

关 键 词：信息处理 嗅敏信息 神经网络 气体检测 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP212[自动化与计算机技术—控制科学与工程]