1997 Volume 59 Issue 6 Pages 83-92
A Kalman neural network (Three layered (7-7-1) neural network) model for quality evaluation of Japanese green tea was developed. Near infrared spectroscopic measurements provided absorbancy data at ten different wavelengths that were correlated to specific ingredients of the tea. Three Colorimetric color values of the green tea specimens were also measured. The principal component analysis (PCA) identified seven suitable input parameters for the network using the spectroscopic and colorimetric data. As the tea quality evaluation was based on the 4 item overall score of standard tea sensory, the 4 item overall score was chosen for the neural network output parameter. The trained neural network performed fairly satisfactorily when tested with inspection data since the correlation factor between actual and estimated data was 0.9290.
