Application and MATLAB realization of drugs' classification based on the combination of NIRS detection and BP ANN algorithm / 国际生物医学工程杂志

ABSTRACT

Objective To realize rapid and non-destructive drug classification and improve the accuracy of drug classification.Methods A model for drug classification based on the combination of principal components analysis and artificial neural network (PCA-ANN) method was introduced.The software for drugs classification was then developed with the utility of MATLAB language.The near infra-red spectrum (NIRS) detection technique was executed on five kinds of drugs (a total of 120 batch samples) and the detection data was collected within the range of 1 350-1 800 nm of excitation wavelength and 0.5 nm of wavelength interval.Results The network training mean square error (MSE) was 5.91e-03,and the prediction error (β) was 2.469％ when the number of the interfering drugs number was less than 5.Conclusions The classification of drugs by NIRS combined with PCA-ANN is feasible and the classification accuracy can be increased.