基于近红外光谱和极限学习机的普洱茶中游离氨基酸总量检测

Free amino acid content detection in Pu'er tea based on near infrared spectroscopy and Extreme Learning Machine

首先用中心化法对光谱数据进行预处理。建立极限学习机模型,其输入层神经元数为1 557,输出层神经元数为1,隐含层神经元数为10。模型得到的训练集均方根误差为0.103 4,预测集均方根误差为0.115 4。结果表明,极限学习机算法能够较准确地检测普洱茶中的游离氨基酸总量。

The spectrum data are pre-dealt with the mean centered method. The Extreme Learning Machine （EI,M） is established with 1 557 input neuron layers, 1 output neuron layer and 10 hidden neurons layers. The root mean square error of calibration set from ELM is 0. 103 4 where the root mean square error of prediction set （RMSEP） is 0. 115 4. The results show that the algorithm can precisely detect the free amino acid content in Purer tea.