Unmixing error of composite spectra as a function of concentration, signal-to-noise ratio, and peak separation

Abstract

We analyze the unmixing error of calculated relative concentration of component spectra as a function of concentration and peak separation. The algorithm is based on least-squares minimization with coefficient optimization. Simulation shows that the error decreases following a power law equation with the increase of concentration and peak separation of component spectra due to the independence of absolute error on concentration. An experiment using different colors of light emitting diodes verifies the result from simulation.