AUTOMATED SPECTRAL SMOOTHING WITH SPATIALLY ADAPTIVE PENALIZED LEAST-SQUARES
Aaron A. Urbas and Steven J. Choquette
Biochemical Science Division, NIST Gaithersburg, MD 20899.
A variety of data smoothing techniques exist to address the issue of noise in spectroscopic data. The vast majority, however, require parameter specification by a knowledgeable user, which is typically accomplished by trial and error. Even when optimal parameters have been determined, though, intrinsic properties of a given smoothing method (e.g. frequency response) can impose limits on achievable noise rejection. The Savitzky-Golay (SG) filter is one of the most popular methods for smoothing and derivative estimation in spectroscopy and is advantageous because of its shape preserving nature and lack of delay. However, in most situations a compromise between noise reduction and signal preservation is unavoidable. For a non-stationary signal, which is representative of many types of spectroscopy data, optimal filter characteristics depend on local signal dynamics and, therefore, adaptive methods are desirable. To address this issue, an adaptive smoothing algorithm based on penalized least squares has been developed, which we term spatially adaptive penalized least-squares (SAPLS). In brief, the method is an iterative optimization procedure that balances data fidelity and roughness until no significant trends, based on multiscale statistics assuming white Gaussian noise, are detected.
Mentors Name: Steven J. Choquette
Division: Biochemical Science
Building/Room/MS: 227 / A159 / 8312
Is your mentor a Sigma Xi Member? No