You are on page 1of 5

Interference Recognition Using

machine learning
Arnav
Aryia
Types of interference
● Narrowband Interference: Spikes or narrow peaks in frequency spectrum , due to other satellites or
devices in same band
● Wideband interference :As continuous noise over large range of frequencies due to spectrum
comms or broadband jammers
● Impulsive interference : sudden, short duration high amplitude , due to lightning or power line
arching
● Co-channel interference:signals from different satellites or ground stations overlap., cause error in
data transmission
● Adjacent channel interference: signals from nearby frequency channels interfere with each other.
● Multipath interference :signals reach by multiple paths due to reflections or refractions, lead to
reduced signal strength
● Jamming : deliberate interference , can be continuous or immediate , severely affect data
transmission
Interferences in bold cause maximum damage and are harder to remove . Jamming is the most disruptive of all
Motivation
It is more preferable to identify the presence of jamming at the initial level, in order to apply proper mitigation

Detection of a jammer is possible at the antenna level by multiple arrays of an antenna and sub-space processing.
This method is good in terms of accuracy, but its complexity, power consumption, and hardware requirement are
more.

At the front end by monitoring automatic gain control of the receiver and noise level of the received signal,
interference can be detected . These techniques require hardware support, but cannot classify the interference.

The most suitable and reliable approach to detect jammer is at pre-correlator stage . The energy level and statistical
parameters monitoring as well as other complex signal processing methods are appropriate at this stage Jamming
detection at the post correlation stage can be applied by monitoring carrier to noise density ratio (C/N0) or
correlation power but mitigation is difficult at this stage.

https://ietresearch.onlinelibrary.wiley.com/doi/full/10.1049/iet-rsn.2019.0091#:~:text=At%20the%20front%20end%20by,%
2Dcorrelator%20stage%20%5B6%5D.
Existing research
● https://www.hindawi.com/journals/mpe/2022/4554374/

Using LSTM , Bi-LSTM and SVM on time-frequency space analysis to predict interference

Deep neural network to predict the value of the signal at a certain time and compare it with the
received signal, using the difference as a feature for interference detection.

● https://scholar.sun.ac.za/server/api/core/bitstreams/0297754f-1965-4c0d-93cf-dae9a97ca4b7/content

This study uses K nearest neighbor and Gaussian mixture models to classify the interference.

● https://ieeexplore.ieee.org/abstract/document/9374638?casa_token=bx6zLwNHqjEAAAAA:S0GOWJ
MQaDQ1_3xOzQbLbCf9EkWRFbsDWNAhqAA-pRfI3i7xBgrORyUPHhUMDwJ-LJj4svBYR5j4

This study has used stacked CNN’s to process a signal and then compares the differences with the
trained data to classify.
Defining the problem
Choosing the right machine learning technique :

● This depends on whether we only want to detect interference or further


classify the type of interference i.e we need to define the output
● For detection a unsupervised ml model could be enough
● While classification will need supervised learning model
● We also need to define the input i.e data
● The structure and complexity of data we have needs to be understood

You might also like