We demonstrate a novel high-resolution neural temperature sensor using two
fiber Bragg gratings (FBG's) and a modular artificial neural network which
is used to learn the mapping relation among frequency, temperature, and the
normalized transmission power spectrum. Because of the fast computational
ability of the modular artificial neural network and the sensitivity of FBG
's, the sensor can make high-resolution temperature and frequency measureme
nts in real time. The experimental results show that the temperature resolu
tion of this sensor can reach 0.005 degrees C.