An adaptive detection system for Raman spectral peaks at low signal‐to‐noise ratios

Abstract

AbstractRaman spectroscopic techniques have made it possible to study adsorbed species on single metal crystals at submonolayer coverages. However, as the amount of coverage decreases, the signal‐to‐noise ratio (SNR) deteriorates. This paper describes an adaptive peak detector (APD), a data analysis system capable of detecting peaks at low SNRs. The basic APD consists of anadaptive linear predictor (ALP) followed by a variance estimator. This detector is robust in that its performance does not rely on a priori knowledge of the exact signal and noise statistics. An experiment was performed in which single Lorentzian peaks embedded in white Gaussian noise at low SNRs were presented to the APD. A dramatic improvement in the detectability of the peaks is demonstrated; the experimental results indicate that the APD is useful at SNRs as low as 0.3. ALP convergence is shown to be a function of the input variance and the length of the predictor. Practical APD implementation considerations are given, and implications for future studies of adsorbed species are discussed.