FPGA-based multispectral fluorometer using CDMA and embedded neural network

We report on the design and implementation of a fluorescence measurement and analysis device that can identify fluorophore substances. The device performs multi-spectral fluorescence measurements, obtained by exciting the unknown substance with light emitting diodes (LED) whose intensity is CDMA coded for noise rejection. The acquired fluorescence data are processed by a CDMA receiver and a neural network for spectral signature identification and measurement. The design's front end exploits the capability of color LEDs to act as photodetectors with different spectral responses when reverse-biased. The system avoids using optical lenses and is implemented as a minimum chip count design where an FPGA performs all of the required processing with the exception of the analog front end. It is thus appropriate for field use.