Noise sources in Raman spectroscopy of biological objects

摘要

We present an overview of noise sources deteriorating the quality of the recorded biological Raman spectra and the ability to determine the specimen composition. The acquired Raman spectra exhibit intense additive noise components or drifts because of low intensity of the scattered light. Therefore we have to apply expensive or bulky measurement setups to limit their inherent noise or to apply additional signal processing to reduce random components after recording the spectra (e.g., Savitzky-Golay filtering, polynomial approximation, denoising by empirical mode decomposition). We present noise sources generated in detectors of Raman scattered photon stream. We consider the methods of background noise reduction by increasing averaging time when the background noise comprises of white noise and 1/f noise components. We consider how the background noise reduces the accuracy of chemical compounds estimation using Raman spectra and prediction model based on linear (e.g., Principal Component Analysis) or nonlinear (e.g., Support Vector Machine) methods. Finally, we give some remarks about synchronous detection and background noise reduction.