拉曼光谱的局域动态移动平均全自动基线校准算法

基线校准是极其重要的光谱预处理步骤，能够显著提高后续光谱分析算法的准确性。目前基线校准算法大多数都是手动或半自动的，手动基线校准算法完全依赖于用户的经验，个人主观因素会严重影响基线校准的准确性，半自动基线校准需要针对不同的拉曼光谱设置不同的优化参数，使用不便。提出了一种局域动态移动平均(LDMA)全自动基线校准算法，并且详细阐明了该算法的基本思想和具体算法步骤。该算法采用了改进移动平均算法(MMA)实现拉曼光谱峰的逐渐剥离，通过自动识别原始拉曼光谱的基线子区间来将整个拉曼光谱区间自动分割为多个拉曼峰子区间，从而实现了在每个拉曼峰子区间中动态改变MMA窗口半宽度和控制平滑迭代次数，最大程度地避免了基线校准过度和基线欠校准现象。无论对于凸形基线、指数形基线、反曲线形基线模拟拉曼光谱，还是真实物质的拉曼光谱，LDMA全自动基线校准算法都取得了很好的基线校准效果。

Locally Dynamically Moving Average Algorithm for the Fully Automated Baseline Correction of Raman Spectrum

The baseline correction is an extremely important spectral preprocessing step and can significantly improve the accuracy of the subsequent spectral analysis algorithm. At present most of the baseline correction algorithms are manual and semi-automated. The manual baseline correction depends on the user experience and its accuracy is greatly affected by the subjective factor. The semi-automated baseline correction needs to set different optimizing parameters for different Raman spectra, which will be inconvenient to users. In this paper, a locally dynamically moving average algorithm (LDMA) for the fully automated baseline correction is presented and its basic ideas and steps are demonstrated in detail. In the LDMA algorithm the modified moving averaging algorithm (MMA) is used to strip the Raman peaks. By automatically finding the baseline subintervals of the raw Raman spectrum to divide the total spectrum range into multi Raman peak subintervals, the LDMA algorithm succeed in dynamically changing the window half width of the MMA algorithm and controlling the numbers of the smoothing iterations in each Raman peak subinterval. Hence, the phenomena of overcorrection and under-correction are avoided to the most degree. The LDMA algorithm has achieved great effect not only to the synthetic Raman spectra with the convex, exponential, or sigmoidal baseline but also to the real Raman spectra.