Research On The Pattern Recognition Method Of Voltammetric Electronic Tongue And Its Application In Food Traceability

The voltammetric electronic tongue is a new type of detection instrument which imitates the mechanism of human taste.It has played an important role in the field of food detection in recent years.However,the existing equipment has the disadvantages of high cost,large size and difficulty in rapid on-site detection.At the same time,the existing pattern recognition method of voltammetric electronic tongue is relatively simple,which greatly restricts the development of voltammetric electronic tongue.Based on the above problems,the main research contents of this paper are as follows:(1)A portable voltammetric electronic tongue system is developed by combining virtual instrument technology with wireless communication technology.The system is mainly composed of a multi-sensor array,a handheld detection terminal,a wireless transmission system,and a host computer terminal.The system has the advantages of small size,easy to carry,etc.,which can meet the needs of rapid and on-field detection for food samples.(2)A voltammetric electronic tongue pattern recognition method based on improved Hilbert-huang transform combined with particle swarm optimization least-square support vector machine was proposed,and the voltammetric electronic tongue system was used to trace the origin of wolfberry from four different producing areas.Firstly,the original signal of electronic tongue was decomposed into a set of intrinsic mode functions by the ensemble empirical mode decomposition.The singular spectral entropy was used to screen the effective intrinsic mode function components,and the singular spectral entropy and Hilbert marginal spectrum were used as feature vectors.On this basis,a linear discriminant analysis model and a least square support vector machine based on particle swarm optimization were used to establish a non-linear combined prediction model of wolfberry origin.The experimental results show that,compared with feature extraction methods such as feature point extraction,principal component analysis,and discrete wavelet transform,the improved Hilbert-huang transform has better classification effect on linear discriminant analysis.The least square support vector machine with particle swarm optimization was used to distinguish the origin of wolfberry.The overall accuracy and Kappa coefficients were 98.5% and 0.98,respectively.(3)A voltammetric electronic tongue pattern recognition method based on adaptive variational mode decomposition-Hilbert spectrum analysis-extreme learning machine was proposed,and five kinds of plant honey were detected by voltammetric electronic tongue.The energy loss function is introduced to make the variational modal decomposition adaptively decomposable.Then,the Hilbert marginal spectrum is obtained as the feature vector in combination with Hilbert spectrum analysis.The area method,discrete wavelet transform,and Hilbert-Huang transform were compared with this method,and the feature extraction effect of proposed method was verified by principal component analysis and extreme learning machine.The experimental results show that the adaptive variational mode decomposition-Hilbert spectrum analysis feature extraction method combined with the principal component analysis can effectively distinguish the five honey samples.This method has a better classification effect than area method,discrete wavelet transform and Hilbert-Huang transform.The training set accuracy,prediction set accuracy,and Kappa coefficient of the pattern recognition method based on adaptive variational mode decomposition-Hilbert spectrum analysis-extreme learning machine were 100%,99.60%,and 0.962,respectively.