My end goal is to create a machine learning model that can classify spectroscopic data from the SparkFun spectral sensor (https://www.sparkfun.com/products/14351). Unfortunately, it's infeasible for me to scan thousands of samples of dirt with the SparkFun sensor in order to create a dataset for training, so what I plan to do is aggregate the spectroscopic signature of DDT with signatures of normal soil samples in order to mimic the signature of a DDT-contaminated soil sample.
The problem is that this resultant data is of MUCH higher resolution than the sensor data, as the sensor data only gives me the intensities as 6 spectral bands as indicated in the manual (https://cdn.sparkfun.com/assets/1/b/7/3/b/AS7263.pdf). I'm not completely sure of the definition of spectral band, but SparkFun describes the bands as "Gaussians centered at wavelength \$\text{X}\$ nm with FWHM of 20 nm" where \$\text{X} \in \{610, 680, 730, 760, 810, 860\}\$
My question is how can I resample my high-resolution spectral data into the format of the SparkFun sensor (i.e, intensities of the 6 spectral bands defined by the sensor).
Additionally, I was also informed that I may need to "normalize" the data that I convert, as the magnitudes of the intensities given by the spectral sensor may not match up with the magnitudes of my converted dataset.
Thank you all!