A FPGA based recurrent neural networks-based impedance spectroscopy system for detection of YAKE in tuna

This paper evaluates the use of impedance spectroscopy combined with artificial intelligence. Both technologies have been widely used in food classification and it is proposed a way to improve classifications using recurrent neural networks that treat the impedance data series at different frequencies as a time series, with the intention of improving the identification of alpha and beta dispersions that are fundamental for the determination of food quality. This proposal in addition to being demonstrated its validity in the detection of YAKE on frozen tuna loins, is fully implemented on a low power FPGA device that allows the classification at the edge by means of a portable equipment that allows its application in food distribution chains with high energy efficiency.