摘要
中药气味的传统嗅觉鉴别往往受到个人鉴别经验和嗅觉能力的限制,客观性和推广性较差。利用电子鼻技术,结合模式识别算法对中药气味进行鉴别分析,可以提高灵敏度与准确度。本研究采用α-FOX3000电子鼻对不同贮藏时间(1年、3年)西洋参进行气味检测,并结合多层感知器网络识别技术建立判别模型;通过十折交叉验证和外部测试集验证对所建模型进行系统性能的评估。另外,采用逐步判别法对传感器阵列进行了优化。结果表明:该模型对不同贮藏时间西洋参具有较高的回判正确率(均为100%)和较好的泛化能力。优化前后的传感器阵列均能实现对不同贮藏时间西洋参的鉴别。为电子鼻在中药研究领域,尤其是中药的"储藏年限"、"有效期"等方面的应用提供实验依据。
The manual identification of traditional Chinese medicine＇s odor is often restricted by the individual ability and subjective experience,thus one can not obtain the precise and objective results.The sensitivity and accuracy of the identification can be significantly improved by the utilization of electronic nose combined with pattern recognition algorithm.α-FOX3000 electronic nose was employed to detect the odor of Radix Panacis Quinquefolii with different storage time（one year three years） while multilayer perception（MLP） network,a novel and efficient feed-forward network,to construct the discrimination model.Firstly,the train set was established.Secondly,a discrimination model was set up and trained based on MLP network.Thirdly,the test set showed that the discrimination model had favorable identification ability and generalization ability.
出处
《中华中医药学刊》
CAS
2013年第7期1683-1685,I0016,共4页
Chinese Archives of Traditional Chinese Medicine
基金
北京中医药大学自主选题项目(JYB22-XS041)
关键词
电子鼻
多层感知器
西洋参
判别模型
electronic nose
multilayer perception
Radix Panacis Quinquefolii
discrimination model
