Browse Prior Art Database

OPTIMIZATION OF PREPROCESSING DESIGN PARAMETERS FOR SPECTROSCOPY

IP.com Disclosure Number: IPCOM000242975D
Publication Date: 2015-Sep-04
Document File: 4 page(s) / 91K

Publishing Venue

The IP.com Prior Art Database

Abstract

A method for optimization of preprocessing design parameters for spectroscopy is disclosed. The method formulates an objective function that is to be minimized. The method selects a technique from optimization techniques generally known in the art. The method obtains a set of raw spectra for a compound of known composition. The method produces mixture spectra and scans the mixture spectra using a spectroscope. The mixture spectra are partitioned into two sets using random sampling. A first set (also called an optimization data set) is used for optimization process and a second set (also called a hold-out set) is used for validation. The method initializes design parameters with values based on a standard practice. The method computes a value of the objective function with initial design parameters on the hold-out set. The method runs an optimizer on the optimization data set. The method evaluates the objective function with optimal design parameters on the hold-out set. If the value of the objective function is smaller as compared to an initial value, the design parameters generalize well. Else, if the value of the objective function increases on the hold-out set, optimization process optimizes against artifacts present in the optimization data sets as opposed to optimizing against real information. In latter case, the optimization process is re-run with stringent termination criterion and results in earlier termination.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 49% of the total text.

OPTIMIZATION OF PREPROCESSING DESIGN PARAMETERS FOR SPECTROSCOPY

BACKGROUND

The present disclosure relates generally to spectroscopy techniques, and more particularly to a method for optimization of preprocessing design parameters for spectroscopy.

There are numerous situations, in which identifying an unknown substance is critical.  Such a situation may be for example, train derailment, overturned vehicles, explosions in chemical plant and presentation of illegal drugs.  

Various conventional analytical methods, such as Raman and Infrared spectroscopy have been used for identifying and characterizing unknown substances.  Raman spectroscopy based techniques usually use a portable instrument with an embedded algorithm for on-site identification of the unknown substances. However, Raman spectroscopy does not perform a preprocessing step to extract spectra of a mixture.  As a result, the spectra are contaminated by noise (i.e. electrical jitter, thermal jitter, cosmic rays) and baseline (i. e. intrinsic fluorescence) issues.  

Preprocessing step improves spectra and extracts a Raman signature of the mixture.  Various methods, such as Savitzky Golay, digital filters, wavelets, polynomial based corrections are used for preprocessing to extract a true signature of the mixture.  However, all such methods require specific design parameters and are very complex.  Further, the conventional methods result in a large number of configurations and evaluating the large number of configurations becomes impractical.

Therefore, it would be desirable to have an improved method for optimization of preprocessing design parameters for spectroscopy. 

BRIEF DESCRIPTION OF DRAWINGS

Figure 1 depicts a flow diagram for optimization of preprocessing design parameters for the spectroscopy.

DETAILED DESCRIPTION

An improved method for optimization of preprocessing design parameters for the spectroscopy is disclosed.  Figure 1 depicts a flow diagram for optimization of preprocessing design parameters for spectroscopy.  

As illustrated in Figure 1, the method starts at step 102 and proceeds to step 104.  At step 104, the method formulates an objective function.  The objective function is to be minimized.  At step 106, the method selects a technique from optimization techniques generally known in the art.  The technique is selected based on the properties of the objective function. The objective function may be discontinuous, continuous, analytic or amenable to computation of gradient analytically etc.  The method proceeds to step 108.

At step 108, the method obtains a set of raw spectra for a compound of known composition.  The method produces mixture spectra either by computer simulation or by physically mixing and scanning the mixture spectra using a spectroscope.  The mixture spectra are partitioned into two sets using random sampling.  A first set (also called an optimization data set) is used for optimization process and a second set (also called a ho...