Unsupervised Target Detection using Canonical Correlation Analysis and its Application to Raman Spectroscopy

We present an unsupervised detection approach, detection with canonical correlation (DCC), for target detection based on a linear mixture model. Our aim is determining the existence of certain targets in a given mixture without specific information on the targets or the background. We use canonical correlations between the target set and the mixed components as the detection index, such that the coefficients of the canonical vector are used to determine the indices of components from a given target library, thus enabling both detection and identification of the components that might be present in the mixture. For applications where the contributions of components are non-negative, we incorporate non- negativity constraints into the canonical correlation analysis framework and derive the corresponding algorithm. We show that DCC and especially its nonnegative variant leads to significant performance gain when applied to detection of surface-deposited chemical agents in Raman spectroscopy.