Identification and localization of rotational spectra using recurrent neural networks
AssigneeUchicago Argonne, LLC
InventorsKirills Prozuments, Daniel P. Zaleski, Prasanna Balaprakash
Life of this patent
Estimating the term…
Abstract
A method of identifying molecular parameters in a complex mixture may include receiving a set of combined transition frequencies and analyzing the set of combined transition frequencies using a first trained artificial neural network to generate a plurality of separated transition frequency sets. Each of the plurality of separated frequency sets may be analyzed using a second trained artificial neural network to generate a respective set of estimated spectral parameters. The method may include identifying a set of molecular parameters corresponding to the set of separated transition frequencies.
Claims20
A computer-implemented method of identifying molecular parameters in a complex mixture, the method comprising:
- receiving a set of combined transition frequencies or an image of a spectrum corresponding to the complex mixture,
- analyzing the set of combined transition frequencies or the image using a first trained artificial neural network to generate a plurality of separated transition frequency sets, and for each set of separated transition frequencies in the plurality of separated transition frequency sets,
- analyzing the set of separated transition frequencies using a second trained artificial neural network to generate a set of estimated spectral parameters;
- and
- identifying a set of molecular parameters corresponding to the set of separated transition frequencies, a set of molecular geometries corresponding to the set of separated transition frequencies, and/or a set of chemical species corresponding to the set of separated transition frequencies.
A complex mixture molecular parameter identification system, comprising:
- a sensor capable of measuring a sample to produce a set of combined transition frequencies corresponding to the complex mixture,
- a user interface via which a user may interact with the complex mixture molecular parameter identification system,
- one or more processors;
- and
- a spectrum analysis application comprising a set of computer-executable instructions stored on one or more memories, wherein the set of computer-executable instructions, when executed by the one or more processors, cause the complex mixture molecular parameter identification system to:
- (i) retrieve, from the sensor, a set of combined transition frequencies, and
- (ii) analyze the set of combined transition frequencies to generate a set of disentangled sets of transition frequencies by analyzing the set of combined transition frequencies using a first trained artificial neural network.
A non-transitory computer readable medium containing program instructions that when executed, cause a computer to:
- receive a set of combined transition frequencies corresponding to a complex mixture, analyze the set of combined transition frequencies using a first trained artificial neural network to generate a plurality of separated transition frequency sets, and
- for each set of separated transition frequencies in the plurality of separated transition frequency sets,
- analyze the set of separated transition frequencies using a second trained artificial neural network to generate a set of estimated spectral parameters, and
- based on the set of estimated spectral parameters, identify a set of molecular parameters corresponding to the set of separated transition frequencies.
Dependencies are inferred from the claim text.