论文部分内容阅读
检测室内有害气体可得到混合有害气体的红外光谱。由于吸收谱带相互交叠的混合气体不易进行定性定量分析,提出了基于特征矩阵联合近似对角化(JADE)的特征提取方法。该方法通过分析数据的高阶统计量信息,充分挖掘原始数据隐含的信息,从而准确区分出混合气体中各物质的光谱,同时应用基于正则理论的支持向量机(SVM)对提取的独立信号源建立多维数据定量分析的模型。实验结果表明,混合气体中各组分的定量分析相关系数均保持在0.999 1以上,验证了该特征提取方法的准确性。
Detection of harmful indoor gases available mixed gas infrared spectrum. Due to the difficulty of qualitative and quantitative analysis of mixed gases with overlapping absorption bands, a feature extraction method based on joint matrices of approximate matrices (JADE) is proposed. By analyzing the high-order statistical information of the data and fully excavating the implicit information of the original data, the method can accurately distinguish the spectra of the substances in the mixed gas, and at the same time, using the support vector machine (SVM) Source The establishment of multi-dimensional data quantitative analysis model. The experimental results show that the correlation coefficients of the quantitative analysis of the components in the mixed gas remain above 0.999 1, which verifies the accuracy of the feature extraction method.