This is a deliberately-simple algorithm and sample library for performing Raman identification.
It is provided as a demonstration of how an external application can be called from Wasatch Photonics ENLIGHTEN™ using the ENLIGHTEN™ plug-in architecture.
In particular, it is the "external application" counterpart of ENLIGHTEN™'s RamanID plug-in, whose source can be found here, and which is included in the ENLIGHTEN™ distribution:
Therefore, although you can test this "application" on its own from the command-line (see Testing, below), it is really intended to be installed alongside ENLIGHTEN™ and called from the RamanID plugin.
It cannot be overemphasized that this is not a "good" Raman ID algorithm. It is provided to demonstrate how an ENLIGHTEN™ user can create their own plugins, even plugins which communicate with external applications at run-time.
More-advanced Raman ID algorithms, with improved "smarts" and accuracy, may be added to ENLIGHTEN™ over time, whether as plugins or integrated features, public or private, but that is beyond the remit and capabilities of this GitHub project.
-
src/
- Files for running identification
-
data/
- Sample test spectra to analyze with the algorithm.
-
libraries/
- Spectral libraries of "pure" compounds. Each library is in its own directory, and contains CSV files representing pure spectra of the named compound.
- WP-785/
- taken from WP-00340
- SiG-785/
- taken from WP-00438
$ make
$ bin/identify --help
I was able to build and run from MinGW with only a couple tweaks, but wasn't sure how to deploy a portable statically linked binary. Therefore I added a Visual C++ project, and currently build on Windows using Visual Studio Community Edition.
You can therefore build this project on Windows by double-clicking the .sln file and compiling under Visual Studio.
The basic goal of this application was to allow the "spectral library" to be nothing more than a folder of spectra saved from ENLIGHTEN™ as individual .CSV files.
This makes it extremely simple for users to generate their own "Raman library" simply by creating a folder and dragging "known-good" spectra into it. Only one spectrum is required (or supported) per compound. Compound names are taken from the "Label" field of the CSV, or filename if that metadata is not provided. By way of demonstration, sample libraries are included in the libraries/ folder.
The algorithm does nothing more than compare peak locations on the x-axis. Peak intensity (relative or absolute) is not factored in the comparison.
Given the above, the peak-finding algorithm in use becomes of paramount importance -- and the peak-finding routine in this application is simple indeed. All the math can be found in \ref Identify::Library, especially:
- \ref Identify::Library::findPeakWavenumbers
- \ref Identify::Library::checkFit
The included sample data and command-line options allow quick testing from a command shell:
$ bin/identify --help
bin/identify Handheld 1.0 (C) 2021, Wasatch Photonics
Usage: bin/identify [--brief] [--verbose] [--alt] [--unknown] [--thresh frac] [--streaming] --library /path/to/library [sample.csv...]
bin/identify --help
NOTE: This version has been modified from the original in the following key respects:
- Only one library compound is supported
- All math is single-precision float
- Memory is static rather than from heap
- Other optimizations as necessary to reduce memory usage and runtime
Example: bin/identify --library libraries/WP-785 data/WP-785/*.csv
Options:
--alt output match 'alternates'
--brief output results on one line per sample
--streaming read streaming spectra from stdin
--thresh unknown threshold (default 0.95)
--unknown output unknown spectral residue
--verbose include debugging output
--auth authenticate only
Example:
$ bin/identify --library libraries/WP-785 data/WP-785/*.csv
Interim: sample 2mmHDPE: matched library 2mmHDPE
Interim: sample 4mmHDPE: matched library 2mmHDPE
Interim: sample Acetone: matched library Acetone
Interim: sample Acetone2mmHDPE: matched library 2mmHDPE
Interim: sample Acetone4mmHDPE: matched library 2mmHDPE
Interim: sample Aklkanes2mmHDPE: matched library 2mmHDPE
Interim: sample Alkanes: matched library Alkanes
Interim: sample Alkanes4mmHDPE: matched library 2mmHDPE
Interim: sample Cyclohexane: matched library Cyclohexane
Interim: sample Cyclohexane2mmHDPE: matched library Cyclohexane
Interim: sample Cyclohexane4mmHDPE: matched library 2mmHDPE
Interim: sample Ethanol: matched library Ethanol
Interim: sample Ethanol2mmHDPE: matched library HDPE-20190905
Interim: sample Ethanol4mmHDPE: matched library HDPE-20190905
Interim: sample Ibuprofen: matched library ibuprofen-20190904-085911-272806-WP-00340
Interim: sample Ibuprofen2mmHDPE: matched library 2mmHDPE
Interim: sample Ibuprofen4mmHDPE: matched library HDPE-20190905
Interim: sample Isopropanol: matched library Isopropanol
Interim: sample Isopropanol2mmHDPE: matched library 2mmHDPE
Interim: sample Isopropanol4mmHDPE: matched library 2mmHDPE
A more accurate algorithm might consider some low-hanging opportunities for improvement:
- interpolate library spectra to match the sample measurement's x-axis (the current design avoids this by only comparing the x-coordinate of peak centroids)
- generate a Pearson correlation coefficient between the sample measurement and each interpolated library spectrum.
- see CHANGELOG