Extraction of Brillouin frequency shift from Brillouin gain spectrum in Brillouin distributed fiber sensors using K nearest neighbor algorithm

The extraction of Brillouin frequency shifts from a Brillouin gain spectrum is a common problem in the engineering applications of Brillouin distributed fiber sensors. In this paper, extraction of Brillouin frequency shifts from Brillouin gain spectrum in Brillouin distributed fiber sensors using K nearest neighbor algorithm is proposed and experimentally demonstrated. The Brillouin frequency shift extraction is treated as a model-free supervised classification problem. The influence of Signal-to-noise ratio and frequency scanning step is studied. The performance of Brillouin frequency shift extraction degrades with the increase of frequency scanning step, and with the decrease of signal-to-noise ratio. The result shows that the algorithm of K nearest neighbor algorithm is a better way to extract the Brillouin frequency shift from Brillouin gain spectrum, comparing with curve-fitting methods, the uncertainty can be 0.44 with frequency scanning step of 1 MHz and SNR of 11 dB.