论文部分内容阅读
在战场等复杂环境得到的混合气体的红外光谱主次吸收峰交错重叠,因此对其定性识别的特征提取方法就显得尤为重要。采集到的各种化学战剂和有机气体的红外光谱数据都是高维度数据,首先采用中心化后降维进行特征提取来尽可能多地捕获到它所包含的本质信息,由于混合气体的红外光谱是非线性、非高斯性信号,把非高斯性作为独立性度量将各成分作为独立分量分离出来,为了满足实时需求,在传统快速独立成分分析(Fast ICA)算法的基础上对其迭代过程进行优化,并应用极限学习机(ELM)建立模型进行定量分析。实验结果表明:改进算法的迭代次数较传统算法减少,定量分析均方差E=2.392 6×10-4,回归系数R=0.999,说明该方法在不影响分离精度的前提下提高了混合物质中纯物质光谱分离出来的效率。
The mixed gas obtained in complex environment such as battlefield has the primary and secondary absorption peaks of the infrared spectrum overlapped and overlapped. Therefore, it is very important to extract the feature of qualitative identification of the mixed gas. The collected infrared radiation spectra of chemical warfare agents and organic gases are all high-dimensional data. Firstly, the feature extraction is performed by using the center-following dimensionality reduction to capture as much as possible of the essential information it contains. Due to the infrared Spectra are non-linear and non-Gaussian signals. Non-Gaussianity is used as an independent measure to separate the components as independent components. In order to meet real-time requirements, the iterative process is based on the traditional fast ICA algorithm Optimize, and use limit learning machine (ELM) to establish a model for quantitative analysis. Experimental results show that the number of iterations of the improved algorithm is less than that of the traditional algorithm, the mean square error of quantitative analysis is E = 2.392 6 × 10-4, and the regression coefficient is R = 0.999, which shows that this method can increase the purity of mixed substances without affecting the separation accuracy The efficiency of material spectrum separation.