cyberquantic logo header
EN-language img
FR-language img
breadcrumbs icon
Telecom

Predictive Analytics

Carrier interference detection and removal for satellite communication

Carrier interference detection and removal for satellite communication

For:
Operators of satellite communication systems Operators of other communication systems (satellite or non-satellite) that are potential sources of interference Users of satellite communication systems Regulation authorities Space agencies
Goal:
Other
Problem addressed
Detection (and possibly classification) of interfering signals in satellite
communication systems (e.g., Digital Video Broadcasting - Satellite - Second
Generation (DVB-S2) or DVB-S2 Extensions (DVB-S2x)), and removal of the
interfering signal using the gained knowledge about the interference
characteristics, with the aim of reducing the error rate at the receiver.
Scope of use case
Machine-learning-based detection, classification and removal of interference
signals for satellite communication systems.
Description
The ML-algorithm operates on the received samples of the
signal consisting of the desired carrier and the interferer.
The ML-algorithm searches for repetitive patterns in the
signal, which are not expected from the known carrier signal.
For instance, the interfering signal can be another DVB-S2 or
DVB-S2x carrier from an adjacent satellite, a radar signal, or
a terrestrial radio relay system. Each of these interfering
signals contains a repetitive pattern, for instance in the form
of pilot symbols or unique words.
Regarding the type of ML method, both supervised and
unsupervised learning can be feasible. However, the
supervised learning scenario requires training using a
number of previously known interferers. This would limit
the detection to a class of selected interfering signals.
The use case can be broken down into different sub-
problems.
Interference detection: This problem can be treated as
anomaly detection, and involves teaching the model
about the undistorted signal from clean data.
Interference classification: Given sufficient training data
on different types of inference signals, the problem can
be treated as a classification problem of undistorted
signals and signals overlapping with a particular type of
distortion. This approach provides the type of distortion
as a result, but may produce unreliable results under the
presence of distortions not trained for. A case that may
necessary to be handled specifically is that of interfering
signals of the same type, e.g., a digital video broadcasting
(DVB) signal overlapping with another DVB signal, as the
statistics of the two signals would be similar, but just the time offset of the synchronization symbols would enable
the identification of the signals.
Signal separation: If interference has been identified,
signal separation may be desirable for further
processing. Parts of the carrier are known (pilot
sequence) or it is possible to transmit known data
signals over the carrier, such that the desired carrier can
be reconstructed at the receiver. The ML-algorithm is
trained using a comparison of the received (and
interfered) signal with the (known) transmitted signal
from the carrier, and determines a model as to how the
interfering samples are added to the carrier. Then the
interference is reduced symbol by symbol from the
carrier based on the trained states of the ML-algorithm.
perceive frame img
Raw Data
understand frame img
Times Series
Interested in the same or similar project?
Submit a request and get a free project evaluation.