基于极值统计的拉曼光谱信噪比评估方法与应用

随着近年来便携式光谱仪技术的迅速发展，CCD光谱仪相对于传统光谱仪在光谱收集方式上发生了很多变化：（1）采集到的光谱对信号进行叠加积分，传统信噪比评估方法无法通过单次检测获得探测器波动；（2）对于谱图噪声（谱线随机波动），由探测器响应随机波动和扫描重复误差转变为CCD探测器像素响应差异、探测器随机噪声和与光学系统分辨力有关的模式噪声。噪声类型发生改变，导致原有的光谱质量评价方法适用性变差，基于实测光谱提出更具适应性的光谱质量评价方法具有很强的现实意义。根据拉曼光谱仪检测器的变化，对采集光谱信号的成分进行分析，在该分析的基础上提出了CCD光谱仪的噪声模型假设，根据该假设使用不同的信号极值点频率对不同的噪声进行像素分离，并对噪声频率模式进行了数值模拟，模拟结果与假设相符；在此基础上提出并实验验证了通过谱线极值间距评估谱线噪声的拉曼光谱信噪比评估方法，该方法包括以下两个步骤：（1）通过采集多次实测光谱进行叠加，叠加过程中对对应不同频次的光谱极值点数量进行统计，得到统计结果后基于文中规律分离光谱仪中的环境噪声和暗噪声；（2）应用上述分离结果，对实测光谱中对应暗噪声的谱线极值点作统计平均，再将该值应用于文中公式，计算得到信噪比。该方法在进行了步骤（1）的前期准备后，可以通过单张谱图评估CCD拉曼光谱仪的随机噪声，并用于评估光谱的信噪比。基于光学构架相同、CCD探测器不同的三个拉曼光谱系统进行实验，采用该方法通过设定信噪比阈值对谱图质量进行控制，获得了一致的光谱曲线；基于该方法对同步叠加平均法进行信噪比拟合，拟合优度达到98%。该方法可用于拉曼光谱仪的性能评估和获取拉曼光谱谱图的质量实时控制。理论和实验表明：对于基于CCD探测器的拉曼光谱仪器，当确定样品和特征峰时，可以基于此方法获得信噪比。该方法还可用于比对不同配置的拉曼光谱设备，以及作为控制谱图质量一致性的标准，并对基于拉曼光谱技术的智能鉴别系统的开发具有指导意义。

Method and Application for Raman Spectra SNR Evaluation Based on Extreme Points Statistics

In recent years, portable spectrometer technology has developed rapidly. Compared with the traditional spectrometer, CCD spectrometer in the spectral collection of the way there have been two changes: (1) The signal is superposed and integrated to generate the spectrum, and the traditional SNR estimation method cannot obtain the detector fluctuation by a single detection. (2) For the spectral noise, the detector responses to random fluctuations and scanning repetitive errors are transformed into differences in pixel response of the CCD detector, detector random noise and mode noise related to the resolution of the optical system. Therefore, it is of great practical significance to propose a more adaptive spectral quality assessment method based on the measured spectrum. According to the changes of Raman spectrometer detector, we analyze the components of the collected spectral signal, and put forward the noise model assumptions of CCD spectrometer on the basis of the analysis. According to this assumption, different signal extremum frequencies are used to separate different noise pixels and the noise frequency mode is numerically simulated. The simulation results are consistent with the assumptions. On the basis of this, we propose and experimentally validate the method for evaluating SNR of Raman spectroscopy to estimate spectral line noise through spectral line spacing. The method includes the following two steps: (1) Collecting multiple measured spectra for superposition, counting the number of spectral extreme points corresponding to different frequencies in the superposition process, and obtaining the statistical results to separate the environmental noise and dark noise in the spectrometer; (2)Applying the above separation results, the statistical average of the spectral line extreme points corresponding to the dark noise in the measured spectrum is calculated, and then the SNR is calculated according to the formula in the text. After the preliminary preparation of step (1), the method can evaluate the random noise of the CCD Raman spectrometer through a single spectrum and evaluate the spectral SNR. In this paper, three Raman spectroscopy systems with the same optical structure and different CCD detectors are used to experiment. By using this method, the spectral quality is controlled by setting the SNR threshold, and a uniform spectral curve is obtained. Based on the method proposed in this paper, the SNR is fitted to the synchronization overlapping average algorithm, and the goodness of fit is up to 98%. The method can be used to evaluate the performance of Raman spectrometer and acquire real-time quality control of Raman spectrum. Theoretical and experimental results show that for the CCD detector-based Raman spectrometer, the SNR can be obtained based on this method when determining the sample and the characteristic peak. The method can also be used to compare different configurations of Raman spectroscopy equipment and as a standard to control the quality of spectra.