Raman spectrum baseline identification and its optimization using genetic algorithm

Laser excitation is an ideal source to obtain Raman spectra by transmitting an incident light and receiving its scattered light. The charge coupled device (CCD) camera is commonly used to capture the low intensity Raman scattered light, which incorporates anti-stokes and stokes scattered light. Fluorescence with certain intensity is inevitably induced by the optic fiber, which acts as one component within the resulting spectrograph. It has a negative effect on the analysis of Raman spectra for sample recognition. The intrinsic Raman spectra extraction is necessary for any application concerning with Raman spectroscopy, especially for the medical applications of real-time human sample diagnosis, such as robotic assisted remote control surgery. Fluorescence backgrounds have very broad spectral features in comparison to Raman features. A suitable baseline needs to be identified, which is capable of actually representing fluorescence. By removing background fluorescence, intrinsic Raman spectra can be obtained for further differentiation by principal component analysis. As the baseline function has a bell shape, the linear polynomial, Lorentzian peak and Gaussian peak spectroscopy functions are used to represent the spectra of fluorescence. For this detrend process to be optimized, genetic algorithm is applied intensively for the global optimization of the baseline simulation. In this research, various approaches are conducted and analyzed for Raman spectrum baseline extraction and an optimal design solution is eventually provided by genetic algorithm computation.