基于时频域特征融合的龙井茶品质判定

Quality Recognition of Longjing Tea Based on Time and Frequency-Domain Feature Fusion

智能感官分析技术是模拟人的感官获取茶叶滋味特征信息,用于茶叶品质自动检测的有效方法。本文采用电子舌智能感官分析仪器采集不同等级西湖龙井茶的智能味觉指纹图谱,从时域和频域两个角度提取电子舌传感器响应信号参数,并将时域特征参数和频域特征参数进行特征融合,然后茶叶样品特征分别采用线性降维方法(主成分分析、线性判别分析)和非线性降维方法(核主成分分析、核线性判别分析)进行维数约简,采用最近邻分类器判定茶叶等级。对单特征参数和多特征参数以及线性降维和非线性降维方法的算法效果进行比较,结果龙井茶等级判定的最高正确识别率在95%以上,实现了电子舌对不同等级龙井茶样品的自动模式分类。

Intelligent sensory analysis could represent the tasting characteristic information by simulating human senses, and it has been applied to tea quality identification successfully. In this paper, electronic nose（E-tongue） was employed for quality classification of Xihu-Longjing tea by collecting the sensing signal information. The signal information was processed from time-domain and frequency-domain independently, and the fusion features with two kinds of features were utilized for dimensionality reduction, by linear dimensional reduction methods（PCA LDA） and nonlinear dimensional reduction methods（KPCA KLDA）. Comparison was conducted between single feature and fusion feature, linear dimensional reduction and nonlinear dimensional reduction, and the highest recognition accuracy achieved higher than95%.