小波变换和神经网络用于红外光谱定量分析

为扣除开放光路法红外光谱中的背景干扰 ,发挥开放光路法的优势 ,提出了针对多组分气体体系的小波变换(WT)和人工神经网络 (ANN)相结合的红外光谱定量方法。该方法在数据处理阶段利用小波变换方法扣除了样品光谱中的背景干扰 ,然后通过计算谱峰强度和组分浓度之间相关系数的方法确定了待测组分的特征峰 ,最后利用 ANN技术实现了定量分析。该方法用现场实测光谱进行了检验。结果显示其综合性能优于其它几种常用的方法 ,是一种有效的红外光谱定量方法

Quantitative analysis of the infrared spectrum using wavelet transforms and artificial neural networks

The background disturbances in the spectra measured using the open path sampling method can be minimized with a new method combining wavelet transforms (WT) and Artificial Neural Networks (ANN) for the quantitative analysis of multi component gas systems. This method mathematically removes the background disturbances during data analysis. Then the feature peaks of the component spectra were determined through calculating the correlation coefficients between the peak strengths and the component concentration. Finally, the non linear relationships between the component concentrations and the peak strengths were obtained using the ANN technique. This method was tested using field measured spectra. The results showed that the system performance was better than with several other commonly used conventional methods. Therefore, the new method can provede effective quantitative analysis of the infrared spectrum.