摘要
利用6只TGS传感器组成的阵列对4种常见的易燃液体和3种不可燃饮料进行测试,并选用4种有代表性的特征提取方法,主元分析法(PCA)、Fisher判别法(FDA)、自组织映射(SOM)、Sammon映射法(Sammon map)作为数据预处理方法,并用3种模式识别方法对预处理后的数据进行识别。结果表明:在各种特征提取方法的处理下,可燃类和不可燃类样本都能被准确地区分,而只有在有导师的特征提取方法才能有效地识别各个可燃液体类子类和不可燃液体类子类的样本类别,最佳的投影维数与各特征提取方法有密切联系,而最优的模式识别方法则与数据的分布有关。
Six TGS sensor array used to identify four common flammable liquids and three kinds of drink with four typical feature extraction techniques and three pattern recognition methods are reported. The four feature extraction techniques involve principle component analysis （PCA）, Fisher discriminant analysis （FDA）, self-organizing mapping （SOM） and Sammon map. The result shows that the flammable species and the incombustible species can be well distinguished under each of the four feature extraction methods ,while the species of samples can only be well identified under the method of FDA. The best dimension of object space depends on the feature extraction techniques,while the best pattern recognition technique has close relationship with the data sets.
出处
《传感器与微系统》
CSCD
北大核心
2007年第8期108-110,113,共4页
Transducer and Microsystem Technologies
关键词
易燃液体
电子鼻
特征提取
降维
模式识别
flammable liquids
electronic nose
feature extraction
dimensionality reduction
pattern recognition
