University of Twente Student Theses

Login

Automatic detection of outlandish signal behaviour in the spectrum of cellular networks

Bakker, Wouter (2019) Automatic detection of outlandish signal behaviour in the spectrum of cellular networks.

[img] PDF
13MB
Abstract:Mobile telecommunications are becoming increasingly important for our society. It is widely used for applications like WhatsApp messaging and social media like Facebook or Instagram. Besides, mobile telecommunications enable vital societal functions and critical infrastructure. Examples of this are calling the emergency services at any time and controlling important infrastructures. A large part of the data in these services is transmitted over cellular networks run by network operators. The Dutch Radiocommunications Agency has a monitoring network to measure all the spectrum usage by these network operators. In this thesis the data from this monitoring network is analysed, and a methodology is presented to automatically detect outlandish signal behaviour in the frequency bands reserved for cellular networks. Besides the detection of outlandish signals, e.g., like interference, the proposed algorithm is also able to detect other events in the spectrum of cellular networks. For example, events that can be detected are a power outage of a base station, power saving at night and changes in aggregated transmit power, e.g., due to deployment of new base stations. The outlandish signal variation detection is basically the detection of an ”unexpected” decrease or increase of the aggregated received power in the network at the measurement points. To perform a structured analysis of the measurement data, a model of the measured signal is proposed. The model describes the relation between the measured signal strength, the wanted signal, the outlandish signals and the noise in the measurement. Furthermore, four cases considering various combinations of the wanted signal, the outlandish signals and the noise in the measurement. The data set was gathered from 15 fixed measurement locations of the Radiocommunications Agency of the Netherlands and contains spectrum data between 20 MHz and 3 GHz. The proposed algorithm has four main steps that lead to the automated detection of outlandish signals in the frequency bands. First, for all 15 locations, noise floors need to be estimated to detect the different channels that are in use by different network operators. This is done by using the median forward consecutive mean excision algorithm (MED-FCME). Second, based on the noise floor threshold, channels in which network operators are active are detected. Then, statistics like the mean, median, minimum maximum and the CDF of all channels are computed. By also computing the variation of the multiple statistics, outlandish signals are detected. Finally, some of these outlandish signals are classified as events when the algorithm is able to connect certain characteristics to the outlandish signal. The whole data collection and detection process is automated with MATLAB. The process is able to combine and automate the data collection, data processing, noise floor detection, event detection and the outlandish signals level detection. It is concluded that the detection process works and that it was possible to detect outlandish signals and events in the downlink channels. The measured values in the uplink bands are too close to the noise floor and were therefore not possible to do a useful analysis on. It was also concluded that the detection of channels only works if there were enough measured values that contain only noise. If the number of measured values that contain noise are relatively small compared to the total amount of measured values, the proposed algorithm is not able to correctly estimate the noise floor. Recommendations are given to increase the success of the proposed algorithm for a next stage. For example, it is possible to detect narrow band signals, if the used measurement bandwidth is decreased. Also, placing measurement antennas inside crowded places like large train stations will increase received signal strength from mobile devices. This enables the possibility to analyse the uplink signals as well. The third recommendation tries to improve the channel detection algorithm by taking into account the local minimum, local maximum and the difference between these two. Finally, the algorithm can also be further improved by doing a deep-dive into improving the way events are currently classified.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:33 physics
Programme:Electrical Engineering MSc (60353)
Link to this item:https://purl.utwente.nl/essays/79013
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page