Data augmentation generator

Data augmentation generator is a data augmentation solution to increase the performance of a classification model by generating realistic data augmented samples.

Usage:

synsamples = data_augmentation_generator( sample1, sample2, per, yaxis_intensifier, xaxis_shift_samples, wavelengths, gaussiansnr, gaussian_samples)

Input:

sample1 - Main sample spectrum (Array 1xN, N is the number of wavelengths/wavenumbers)
sample2 - Second sample spectrum for the blender block (Array 1xN)
per - concentration grades for the blender block (Array)
yaxis_intensifier - amplification factor for spectral intensifier
xaxis_shift_samples - number of samples produced by shifting along x axis block
wavelengths - wavelengths/wavenumbers of the spectra needed for the shifting along x axis block (Array 1xN)
gaussiansnr - Gaussian noise signal-to-noise ratio per spectrum, in dB
gaussian_samples - number of samples produced by adding noise block

Output:

synsamples - the generated spectra.

Examples:

1. This example makes use of the blender block where sample1 and sample2 are used for generating artificial admixtures for per different concentration grades (weighted sum of the input samples).

per = [0.16,0.18,0.20,0.40,0.60,0.70,0.80,0.82,0.84];
new_samples = data_augmentation_generator(sample1,sample2,per,0,0,[],0,0); 

2. The following example generates 63 data augmented samples of sample1 by applying spectral intensifier: M= 1.01 to 1.33 with a step of 0.01, shifting along x-axis: Laplacian distribution with b = 0.6 (20 samples), and noise: 10 samples with 35 dB Gaussian noise.

new_samples = data_augmentation_generator(sample1,[],[],33,20,wavelengths,35,10);

If you use data augmentation generator we would appreciate a citation to:

Georgouli, K., Osorio, M.T., Martinez Del Rincon, J. and Koidis, A., 2018. Data augmentation in food science: Synthesising spectroscopic data of vegetable oils for performance enhancement. Journal of Chemometrics, 32(6), p.e3004.