Figure 5.

Feature extraction and pattern recognition of detected ultraweak photon emission (UPE) by a chip composed of complementary metal-oxide-semiconductor (CMOS) array via convolutional neural network (CNN) on a software installed on a computer, machine, or smartphone; (top) direct UPE detection without optical interferometer, and (bottom) UPE detection after the interferometer. The existence of optical interferometer is to discriminate UPE wavelengths, since interference of similar photons (in wavelength) make a different pattern with non-similar photons. One of the advantages of such an interferometer is to have a simple “spectrometry” over similar wavelengths. However, an ensemble of wavelengths may make different patterns at the same time and obscure the interference patterns which may not make advantage over a direct detection, but one can classify those ensemble patterns with pattern recognition techniques such as PCA, which can find the differences between different patterns in the overlapped patterns, and classify each pattern for the relevant wavelength after many sets of training data. The direct detection of UPE by CMOS array and indirect detection after an optical interferometer both can be used for UPE data acquisition.