为了提高拉曼光谱检测系统的时间分辨率, 常常需要采用较短的采样积分时间, 此时带有分子结构振动谱的有用拉曼信号可能完全淹没在噪声中, 严重影响信号的进一步分析, 因此有必要对测量所得的光谱信号进行噪声消除处理。 传统的消噪方法是基于信号与噪声在频域或统计特性之间的差异, 通过平滑滤波或取平均值的方法来消除噪声, 一般适用于噪声强度不高的情况, 对于信噪比较低的情况处理效果并不理想。 针对传统去噪方法的不足, 从信号重构的角度, 利用基于小波变换的谱峰识别、 半峰宽检测提取光谱特征参数, 再利用最小二乘拟合的方法, 能够有效地提取淹没于强噪声背景下的有用拉曼信号。 在仿真中, 运用该算法得到的光谱曲线光滑, 峰位置准确, 信噪比改善明显。 在实验中, 分别利用该方法处理头孢呋辛酯片和罗红霉素拉曼光谱数据, 得到了清晰的谱峰位置、 幅值及半峰宽信息, 实现了对短积分时间、 强噪声背景的拉曼信号的有效还原, 提高了检测系统的时间分辨率。 仿真和实验结果表明, 该方法需要调整参数少, 易于实现, 在信噪比比较低的情况下依然能够得到良好的去噪效果, 为进一步分析光谱数据提供准确可靠的信息。

To improve time resolution of the Raman measurement system, we need to adopt short scanning time. In this case, the weak Raman signal with vibrational spectrum of the molecular structure is easily to be buried by the high background noise, which influences the further analysis seriously. So it is necessary to de-noise the raw Raman signals. Conventional methods manage to de-noise signal by means of smoothing or averaging based on the difference between signal and noise in frequency characteristic or statistical features. They are commonly applied in the situation where the background noise is not so strong, and cannot give satisfactory results to the Raman signals with low signal-to-noise ratio. In this paper, the algorithm proposed detects peak positions and get peak half-width based on wavelet transform, and then reconstructs the Raman signals by least square fitting algorithm with characteristic parameters obtained, which extracts the useful signal from high background noise efficiently. In the simulation, the Raman curve fitted by the proposed algorithm was smooth, and the peak positions obtained were accurate, so the signal-to-noise ratio improved significantly. In the experiment, we adopted this algorithm to de-noise the tested Raman signal of Cefuroxime Axetil Tablets and Roxithromycin, respectively. The peak positions, peak half-width and amplitude were obtained and proved to be accurate. Therefore, the useful pure Raman signal could be recovered from the high background noise efficiently by the proposed algorithm, which improved the time resolution of Raman system. Both the simulation and the experiment showed that the proposed method could be easily performed with only a few parameters. Comparing with conventional methods, it could achieve satisfactory results under high background noise and provide accurate and reliable information for further analysis.