The aim of this paper was to explore the non-destructive application of fusion spectra model for characterization and differentiation of electrostatic copying papers that could be favourable to give forensic aid in legal cases. Two hundred fifty electrostatic copy paper samples were collected from various markets. All samples were subjected to surface-enhanced Raman spectroscopy (SERS) analysis from 3500 to 400 cm⁻¹ Raman shift range and attenuated total reflection-Fourier transform infrared spectrum (ATR-FTIR) analysis from 4000 to 400 cm⁻¹ wavenumber range, respectively. The spectral data refracted the constituents present in the electrostatic copy papers were cellulose, inorganic filler calcium carbonate and barium sulfate. Fisher discriminant analysis (FDA), multi-class support vector machine (MSVM) and decision tree (DT) algorithms were used to build the model. The precision rate, recall rate, F-score and total accuracy were considered as indicators to evaluate the model's performance. The results showed that fusion models were superior to single model. The feature layer fusion model based on MSVM algorithm gave a differentiating power of 100% by grouping all the sample groups. This study demonstrated that spectral fusion model is a feasible and reliable approach for fast and non-destructive differentiation of electrostatic copying papers. It has great potential in real application scenarios.

中文翻译：

---

在数据融合模型中结合表面增强拉曼光谱和中红外光谱来法医区分静电复印纸

本文的目的是探索融合光谱模型在静电复印纸表征和区分方面的无损应用，以利于在法律案件中提供取证援助。从各个市场收集了 250 份静电复印纸样品。所有样品分别在3500至400 cm ^{-1拉曼位移范围内进行表面增强拉曼光谱（SERS）分析，并在4000至400 cm -1}波数范围内进行衰减全反射-傅立叶变换红外光谱（ATR-FTIR）分析。。光谱数据折射出静电复印纸中存在的成分是纤维素、无机填料碳酸钙和硫酸钡。使用费舍尔判别分析（FDA）、多类支持向量机（MSVM）和决策树（DT）算法来构建模型。将准确率、召回率、F值和总准确率作为评估模型性能的指标。结果表明，融合模型优于单一模型。基于MSVM算法的特征层融合模型通过对所有样本组进行分组，给出了100%的区分能力。这项研究表明，光谱融合模型是快速、无损区分静电复印纸的一种可行且可靠的方法。在实际应用场景中具有巨大的潜力。