Determination Of Sodium Cyclamate Content In Jelly By Near Infrared Spectroscopy

Sodium Cyclamate is a kind of sweet food additive, which is commonly used in jelly .Excessive usage of it will do harms to human health. It has been clearly stated by the government in national standard GB2760-1996《Food Additive Hygeian Standard》that the Sodium Cyclamate must be less than 0.65g/kg in jelly. However current method for Sodium Cyclamate measurement is time consuming; with complicated procedures to follow, the cost is remarkably high; besides, the accuracy is doubtful since that it is affected by too many factors. Near-infrared (NIR) spectroscopy analysis is one kind of frequently used rapid and non-destructive measuring technique in the measurement of food and farm products. Our research focuses on using NCR spectroscopy analysis technique to measure the content of Sodium Cyclamate in jelly.A set of 66 "Xizhilang" brand jelly is selected as samples in our work, 36 of which are used as a training set, the remaining 30 are as a prediction set. Ultraviolet Absorption Spectrometry (UAS) is adopted for comparing purpose; first we use UAS to measure the content of Sodium Cyclamate in the jelly, afterwards we collect the Near-infrared spectral data of the same set of jelly; Most part of our work is based on the data collected from both kinds of experiment data. Multiple preprocessing methods are utilized to process the Near-infrared spectral data obtained, that include Savitzky-Golay smoothing, derivation, standard normal variate transformation, multiplicative scatter correction, mean-centering, wavelet transform and so on.; The processing results are also briefly compared. To select appropriate wavelength, we have analyzed the variance of the possible spectral interval from 4000～11000cm^-1 we find that the interval of 4001.64～6998.52cm^-1 is closely related to the content of Sodium Cyclamate. Furthermore we have used both Intervals Partial Least Square (iPLS) and Genetic Algorithm (GA) to identify a tighter range of wavelength, by comparing the results of both methods, it is clear that GA have a better performance. Finally a model of Artificial Neural Network (ANN) with Principal Component Analysis (PCA) is constructed and compared with iPLS Model; by comparison, we select the model of better performance as our model, i.e. a ANN model based on the wavelength selected by GA; the Root Mean Square Error of Cross-validation (RMSECV) of the model is 0.00014, Root Mean Squared Error of Predication is 0.00093, the Relation Coefficients are 0.9173 and 0.9056 , The result is rather satisfactory.Results show that Near-infrared (NIR) spectroscopy analysis can be used to measure the content of Sodium Cyclamate in jelly. The research work on spectra collection, pre-processing,wavelength selection and modeling that have been done in this paper will provide valuable references for future researching work in this aspect.