IEEE - Aerospace and Electronic Systems - June 2022 - 33

Image licensed by Ingram Publishing
A remote ground station equipped with a cognitive radio
requires the operator to do little more than turn ON the power,
set the carrier frequency, align the receive antenna to the test
article, and push " go. " Second, misconfigurations can be
avoided. It can sometimes be the case that the flight test engineer
in the telemetry ground station does not know which of
the three modulations is being used by a test article. This can
happen when incorrect information is given to the flight test
engineer or ifa multimode transmitter has been installed with
the incorrect settings. Third, a handheld version of the cognitive
radio can be used by flight-line technicians tomonitor the
output ofa test article and confirm proper configuration ofthe
telemetry downlink before the test article goes airborne.
The idea is illustrated in Figure 1. The cognitive radio
comprises two elements: the machine learning algorithms
and the software-defined radio (SDR). The role of machine
learning is to determine two parameters: the modulation type
and the bitrate. An SDR equipped with these two pieces of
information is capable of demodulating and detecting the
telemetry information bits. The modulation type is required
because the optimal detectors for the three CPMs are different.
The bitrate is required to configure the resampling filters
to produce an (approximate) integer number of samples/bit
for the symbol timing synchronizers.
The ability to determine modulation type is called modulation
recognition or classification. Automatic modulation
recognition for linear modulations is a relatively well-known
problem. See the survey [9] and the references therein. The
modulation recognition problem for CPM is more difficult
and, for the most part, has been devoted to recognizing the
difference between a linear quadrature amplitude modulation
(QAM) and simple frequency shift keying (FSK), a special
case of full response CPM [9]. An exception to this is the
work described in [10] wherein entropy is used as a classifier
for different partial response CPMs. However, the work
described in [9] and [10] assumes that the bitrate is known.
Because this assumption cannot be made in aeronautical
mobile telemetry, direct application of the techniques in [9]
and [10] does not work here. For this reason, the use of
machine learning is investigated.
In this article, we apply machine learning to the modulation
recognition and bitrate estimation problems. The results
show that a quadratic discriminant classifier trained on the
signal power spectral density (PSD) estimate is best suited to
identifying the modulation type and a Gaussian process
regression algorithm is best suited to determining the bitrate.
PRELIMINARIES
PCM/FM has been used since the 1970s and is a partial
response CPM with a binary alphabet, modulation index
h ¼ 0.7, and a frequency pulse spanning three bit times.
PCM/FM is the baseline for the modulations that were
adopted later. The purpose of the ARTM program, established
in 1997, was to improve the spectral efficiency of the
RF link. This goal was accomplished in two stages or " tiers " :
the first was to increase the spectral efficiency by a factor of
2 relative to PCM/FM and the second was to increase the
Figure 1.
Concept of a cognitive radio for aeronautical telemetry.
JUNE 2022
IEEE A&E SYSTEMS MAGAZINE
33
IEEE - Aerospace and Electronic Systems - June 2022

Table of Contents for the Digital Edition of IEEE - Aerospace and Electronic Systems - June 2022
