摘要
针对现有电子鼻系统训练误差大、运行速度慢等特点,提出了一种新的基于在线支持向量机(Online-SVM)的电子鼻系统模式识别方法。该方法使用CH4气体与传感器阵列响应的值作为输入数据,经在线支持向量机算法进行模式识别,对CH4气体的浓度进行预测和分类。与期望结果对比,新方法的平均误差降低为5.3%,运行时间降为0.1994s,表明基于在线支持向量机的电子鼻系统模式识别方法能有效提高电子鼻系统识别的精度和速度。
A new online support vector machine （online-SVM） method was presented as the electronic nose system pattern recognition algorithm in the paper to solve the problems of big training error and low execution speed which exist in current electronic nose systems. In the paper, the response values from CH4 gas sensor array was the input data, and online support vector machine algorithm was used as the pattern recognition algorithm to predict and classify the concentration of CH4 gas to obtain predictions. Compared with the expected results, the average error of the proposed algorithm was lowered to about 5.3%, and the operating time was reduced to 0. 199 4s, which showed that the new algorithm based on online-SVM could improve the discrimination accuracy and execution speed of the electronic nose systems.
出处
《西北大学学报（自然科学版）》
CAS
CSCD
北大核心
2015年第1期49-52,共4页
Journal of Northwest University（Natural Science Edition）
基金
四川省教育厅重点项目资助(14ZA0286)
四川省应用基础研究计划资助项目(2013SZZO)
云南省应用基础研究计划基金资助项目(2011FZ037)
关键词
电子鼻
在线支持向量机
模式识别
预测
损失函数
electronic nose
online support vector machine
pattern recognition
forecasting
loss function
