Intelligent control application on sample identification

An intelligent control implementation is proposed for sample differentiation with Raman spectroscopy, which can be used to characterize various samples for decision-making and medical diagnosis. Raman spectra are weak signals whose features are inevitably affected by numerous noises during the calibration process. These noises must be eliminated to an acceptable level. Fuzzy control has been widely used to solve uncertainty, imprecision and vague phenomena, so fuzzy logic can be used for noise filtering. The resulting intrinsic Raman spectrum has been trained using artificial neural networks. Both unsupervised learning and supervised learning are to be conducted in this preliminary research on sample identification. For unsupervised training, principal component analysis (PCA) is exploited, which is based on Hebbian rule and single value decomposition (SVD) approach, respectively. For supervised training, radial basis function (RBF) is presented. A complete procedure for sample identification consists of Raman spectra calibration, noise filtering, unsupervised classification and supervised neural network training. A systematic intelligent control approach is formulated in consequence for sample identification. The long-term objective is to create a real-time approach for sample analysis using a Raman spectrometer directly mounted at the end-effector of the medical robots to enhance robotic remote surgery