Research On Food And Drug Analysis Based On Electronic Tongue And Pattern Recognition Technology

With the rapid economic development and the improvement of people's living standard in our country,the demand for food,drug quality detection and control is growing rapidly.It not only requires the detection method to be accurate and objective,but also requires easy operation and rapid detection.Electronic tongue technology is a new detection technology,which detects the whole "fingerprint" information of liquid sample through sensor array,and combines the pattern recognition methods to achieve qualitative and quantitative analysis of the tested sample.In this paper,a set of voltammetric electronic tongue system based on virtual instrument technology has been developed by our group.Based on this system,taking the Chinese patent medicines,freshly squeezed orange juice and pericarpium citri reticulatae as the research object,an applied research was performed on the suitability of the system in food and drug quality detection and control,in which the different pattern recognition methods were comparatively explored.The specific researches are as follows:(1)A voltammetric electronic tongue system based on virtual instrument technology was developed.The system consists of four parts: sensor array,signal conditioning module,data acquisition card and upper computer software.It has the advantages of small volume,low cost,rapid detection,objective and efficient,et al.(2)4 kinds of Chinese patent medicines in the treatment of cold symptom were analyzed by the voltammetric electronic tongue.The feature extraction program of electronic tongue output signal was first performed by the feature point extraction method and discrete wavelet transform method,respectively.According to clustered property and classification effect of sample points,the discrete wavelet transform was selected as a recommended feature extraction method.The principal component analysis,cluster analysis and back propagation neural network were then used to distinguish and identify different kinds of Chinese patent medicines.The results showed that back propagation neural network exhibited a better result.(3)The fresh orange juice with different storage time was analyzed by the voltammetric electronic tongue.According to the characteristics of electronic tongue respond signal,the principal component analysis method and discrete wavelet transform method was first used to feature extraction of respond signal,respectively.According to the classification result,the discrete wavelet transform was selected as a recommended feature extraction method.Then the linear discriminant analysis was used to the qualitative analysis of fresh orange juice samples with different storage time.And the least squared-support vector machines based on particle swarm optimization method was applied to quantitative forecast the different storage time.The results showed that the system can effectively distinguish samples of freshly squeezed orange juice at different storage times and quantitatively predict the storage time of freshly squeezed orange juice.(4)The voltammetric electronic tongue was attempted to classify pericarpium citri reticulatae(PCR)according to different ages for authentication.The characteristic variables were extracted from the responses of the sensors by discrete wavelet transform.Seven pattern recognition methods,which were principal component analysis,linear discriminant analysis,cluster analysis,back-propagation neural network,support vector machine,random forest and extreme learning machine,were compared for developing the discrimination model of PCR with different storage ages.Experimental results showed that extreme learning machine model exhibited the best performance in the classification of PCR with different storage ages.