IR spectra matching methods
Spectra matching is widely used in various applications including the search for a spectrum of an unknown or subject material, chemical, or compound in an existing spectral database and quality control by means of comparing the spectra of products with standards. New systems and methods are described for identifying an unknown compound by calculating the similarities of Fourier-transform infrared (FTIR) spectra of organic compounds. The systems and methods incrementally calculate the spectral similarity based on the local spectral shapes. This reduces the bias caused by uneven weighing of large or broader peaks. In addition, the new systems and methods tolerant to the common issues in spectra matching including baseline offset, baseline sloping, and deviations in wavenumber axis alignment, suggesting its robustness and practical applicability.
1. A method for identifying a subject material having a subject spectrum, the method comprising:
obtaining the subject spectrum for the subject material by measuring spectral data associated with the subject material;
obtaining from a database spectral information associated with a plurality of reference spectra for a plurality of reference materials;
calculating a final set of normalized local change values for the subject spectrum by:
determining a set of ranges for each discrete wavenumber value of a plurality of discrete wavenumber values, each set of ranges including an upper wavenumber range and a lower wavenumber range;
determining for each discrete wavenumber value a sum of absorbance data values of respective upper and lower wavenumber ranges; and
calculating for each discrete wavenumber value a normalized local change value by dividing the sum of the absorbance data values of the respective upper wavenumber range by the sum of the absorbance data values of the respective upper and lower wavenumber ranges,
wherein each upper wavenumber range extends between respective first and second upper wavenumber values, each of the first and second upper wavenumber values being higher than the respective discrete wavenumber value,
wherein each lower wavenumber range extends between respective first and second lower wavenumber values, each of the first and second lower wavenumber values being lower than the respective discrete wavenumber value;
calculating a final spectral similarity value for each reference spectrum in the plurality of reference spectra by comparing a final set of normalized local change values for each reference spectrum with the final set of normalized local change values for the subject spectrum;
comparing the final spectral similarity values to determine a highest spectral similarity value; and
identifying the subject material as the reference material associated with the highest final spectral similarity value,
wherein each normalized local change value of each final set of normalized local change values is between zero and one.
2. The method of claim 1 , wherein each first upper wavenumber value is one unit greater than its respective discrete wavenumber value of the plurality of discrete wavenumber values, wherein each second upper wavenumber value is at least one unit greater than its respective first upper wavenumber value, wherein each second lower wavenumber value is one unit less than its respective discrete wavenumber value of the plurality of discrete wavenumber values, and wherein each first lower wavenumber value is at least one unit less than its respective second lower wavenumber value.
3. The method of claim 1 , wherein the step of calculating the final set of normalized local change values for the subject spectrum further comprises:
selecting an optimized range value; and
using the optimized range value to determine for each of the plurality of discrete wavenumber values respective first and second lower wavenumber values and respective first and second upper wavenumber values.
4. The method of claim 3 , further comprising determining the optimized range value, wherein determining the optimized range value comprises:
selecting a first reference spectrum of the plurality of reference spectra, the first reference spectrum being a first optimizing spectrum;
calculating a first set of normalized local change values for the first optimizing spectrum;
calculating a first spectral similarity value for each of the reference spectrum in the plurality of reference spectra by comparing a first set of normalized local change values for each reference spectrum with the first set of normalized local change values for the first optimizing spectrum, wherein each normalized local change value associated with each first set of normalized local changes values is calculated using a first range value;
calculating a plurality of subsequent spectral similarity values utilizing a plurality of subsequent range values; and
calculating a success rate and an average hit index for the first range value and for each of the plurality of subsequent range values.
5. The method of claim 1 , wherein the step of calculating the final set of normalized local change values for the subject spectrum further comprises:
determining an optimized floor value; and
replacing all absorbance data values in the subject spectrum that are less than the optimized floor value with the optimized floor value.
6. The method of claim 5 , wherein determining the optimized floor value comprises:
calculating an average absorbance value of the absorbance data values associated with the subject spectrum; and
multiplying the average absorbance value by an optimized floor multiplier value.
7. The method of claim 6 , wherein the optimized floor multiplier value is determined by:
selecting a first reference spectrum of the plurality of reference spectra, the first reference spectrum being a first optimizing spectrum;
calculating a first spectral similarity value for each reference spectrum in the plurality of reference spectra by comparing a first set of normalized local change values for each reference spectrum with a first set of normalized local change values for the first optimizing spectrum, wherein each normalized local change value associated with each first set of normalized local changes values is calculated using a first floor multiplier value;
calculating a plurality of subsequent spectral similarity values utilizing a plurality of subsequent floor multiplier values; and
calculating a success rate and an average hit index for the first floor multiplier value and for each of the plurality of subsequent floor multiplier.
8. The method of claim 1 , wherein calculating the final spectral similarity value for each reference spectrum comprises:
associating a first wavenumber value from the subject spectrum with a first wavenumber value from a first reference spectrum of the plurality of reference spectrum; and
calculating a first spectral similarity value using a first normalized local change value associated with the first wavenumber value from the subject spectrum and a first normalized local change value associated with the first wavenumber value from the first reference spectrum.
9. The method of claim 8 , wherein associating the first wavenumber value from the subject spectrum with the first wavenumber value from the first reference spectrum comprises:
selecting the closest wavenumber values from the subject spectrum and the reference spectrum that are associated with an absorbance value; and
associating the selected closest wavenumber values.
10. The method of claim 9 , wherein the wavenumber values are selected only if the difference between the wavenumber values is less than or equal to a predetermined maximum difference value.
11. The method of claim 8 , further comprising repeating the step of associating a wavenumber value for a plurality of wavenumbers in the reference spectrum.
12. A method for identifying a reference spectrum for an unknown subject spectrum from among a plurality of reference spectra, the subject spectrum and each reference spectrum comprising a plurality of discrete wavenumber values, each discrete wavenumber value being associated with respective upper and lower wavenumber ranges, each range being associated with a plurality of absorbance values, the method comprising:
obtaining the subject spectrum by measuring spectral data associated with a subject material;
calculating a final set of normalized local change values for the subject spectrum;
calculating a final set of normalized local change values for each reference spectrum in the plurality of reference spectra;
calculating a final spectral similarity value for each reference spectrum using the final set of normalized local change values for the subject spectrum and the final set of normalized local change values for the relevant reference spectrum;
identifying a maximum final spectral similarity value of the calculated final spectral similarity values; and
identifying the reference spectrum for the unknown subject spectrum as the reference spectrum associated with the maximum spectral similarity value,
wherein calculating the final set of normalized local change values for the subject spectrum comprises:
determining for each discrete wavenumber value sums of absorbance values within respective upper and lower wavenumber ranges, wherein each upper wavenumber range consists of absorbance values for wavenumber values that are higher than the respective discrete wavenumber value, and wherein each lower wavenumber range consists of absorbance values for wavenumber values that are lower than the respective discrete wavenumber value; and
calculating for each discrete wavenumber value a normalized local change value by dividing the sum of the absorbance values of the respective upper wavenumber range by the sum of the absorbance values of the respective upper and lower wavenumber ranges.
13. The method of claim 12 , wherein calculating the normalized local change value for each discrete wavenumber value further comprises:
selecting an optimized range value; and selecting an optimized floor multiplier value; and
using the optimized range value to determine upper and lower bounds for each upper and lower wavenumber range.
14. The method of claim 12 wherein the step of calculating a spectral similarity for a reference spectrum and the subject spectrum further comprises pairing a target wavenumber from the set of normalized local change values for the reference spectrum with a target wavenumber from the set of normalized local change values for the subject spectrum.
15. The method of claim 14 wherein the step of calculating a spectral similarity for a reference spectrum and the subject spectrum further comprises calculating a spectral difference value based on the absorbance value for the paired target wavenumbers from the reference spectrum and the subject spectrum.
16. The method of claim 12 , further comprising repeating the step of associating a wavenumber value for a plurality of wavenumbers in the reference spectrum.
17. The method of claim 12 , wherein each upper wavenumber range extends between first and second upper wavenumber values, the first upper wavenumber value being one unit greater than the respective discrete wavenumber and the second upper wavenumber value being at least one unit greater than first upper wavenumber value, and wherein each lower wavenumber range extends between first and second lower wavenumber values, the second lower wavenumber value being one unit less than the respective discrete wavenumber and the first lower wavenumber value being at least one unit less than the second lower wavenumber value.