报道了一种基于卡尔曼滤波的线性神经网络用于 B族维生素混合组分的定量同时测定 ,用相关系数和标准偏差从原始紫外光谱数据中挑选 1 1个波长点供线性神经网络处理 ,以极少的迭代次数达到较高的精度 ,预测结果较好 ,相关系数大于 0 .99983 ,标准偏差 0 .1 1 3 0 1 A linear neural network based on Kalman filter is reported for quantitative simultaneous determination of B vitamins. The correlation coefficients and standard deviations are used to select 11 wavelength points from the original UV spectrum data for linear neural network processing. Very few iterations to achieve high accuracy, the prediction is better, the correlation coefficient is greater than 0.99983, the standard deviation of 0 .1 1 3 0 1