Single-mode optical fibre digital decoder based on polarization using a
K-Nearest Neighbour algorithm
D. Ibáñez-Camarillo1, F.
Martínez-Piñón1, M.V.
Márquez-Olivera1, V.G.
Hernández-Herrera1 and S.
Vidal-Beltrán2
1 Instituto Politécnico Nacional, Centro de
Investigación e Innovación Tecnológica, Azcapotzalco, Ciudad de México,
02250 México.
2 Instituto Politécnico Nacional, Escuela
Superior de Ingeniería Mecánica y Eléctrica, Zacatenco, Ciudad de
México, 07708 México
Email: fmartinezp@ipn.mx
It is shown experimentally a new digital optical decoding scheme based
on the transmission or polarized light at p polarization planes using a
K-Nearest Neighbor (KNN) algorithm through a single-mode optical fibre
at 633 nm. The optical power signal is sent at p polarization planes
which constitute p classes required for signal bit recognition. Results
show that it is possible to recognize 32 polarizations planes, 5 bits,
using 4 features corresponding to the measurement of optical power at 4
different angles at the photodetector side with an average assertiveness
of 99.1%.
Introduction: The use of artificial intelligence has recently
increased in many areas of science and engineering, including wireless
telecommunications , optical fibre communications and optical fibre
sensor applications . In this paper we explore experimentally the use of
the K-Nearest Neighbour algorithm (KNN) for the integration of a fibre
optics digital decoder based on light polarization to estimate 32
angular positions of a device. According with , KNN is a supervised
learning classification algorithm that uses the proximity of the data to
classify with respect to a data base. For this reason, a data base has
been created that stores information about the optical power pattern in
the output of a single-mode optical fibre for each of the 32 transmitted
signals encoded in polarization.
Experiment: Fig. 1 shows the elements required to generate a KNN
database. Light from a He-Ne laser is coupled into a single mode fibre
SM600 (Thorlabs) through a polarizer disk that has the function of
modifying the polarization angle in which the light is launched into the
optical fibre. For the transmission of 5 bits is necessary to control 32
polarization planes in 180° and each bit sequence has a specific
polarization angle associated with it as shown in Table 1. The coding
disk will take any of these 32 positions which are called “classes”.
On the other hand, with the analyzer disk, 4 “features” are
obtained that characterize the optical power pattern received by the
photodiode of the optical power meter at the output of the fibre. These
features are records of the optical power measured for 4 positions of
the analyzer disk (0°, 45°, 90° and 135°) for each one of the classes.
Table 2 shows a section of the data base can be formed with 32 classes,
100 inputs per class and 4 features for each input.