Unit:
Κατεύθυνση / ειδίκευση Διαχείριση Πληροφορίας και Δεδομένων (ΔΕΔ)Library of the School of Science
Author:
Κογχυλάκης Ευάγγελος
Supervisors info:
Ευστάθιος Χατζηευθυμιάδης Επίκ. Καθηγητής
Original Title:
Μονο- και πολυμεταβλητές αυτοπαλίνδρομες τεχνικές για συμπίεση δεδομένων σε ασύρματα δίκτυα αισθητήρων
Translated title:
Uni- and multi-variate autoregressive techniques for data compression in wireless sensor networks
Summary:
Wireless sensor networks are found in numerous applications in most scientific
fields as the collection of information from individual and mobile units
becomes necessary. An integral part of the network of sensors is the actual
transfer of data, which is often critical and for this particular reason there
should be a recovery mechanism in case of information loss. The sensors which
are the building blocks of networks are characterized by the well-known problem
of limited energy resources.The purpose of this paper is to introduce new data
processing methods based on which the relevant information is compressed before
it is sent to the rest of the network. Also the new algorithm has an added
module which is a frame prediction algorithm where the lost information is
being predicted efficiently also used to further compress each frame.Such an
algorithm was designed and implemented. This algorithm combines compressing the
information while retaining most of its significance and the predictability of
the information. Principal component analysis and the vector autoregressive
model were used. The algorithm was compared with other univariate prediction
algorithms using different metrics for different datasets. The results are
promising and in particular the average mean squared error is satisfactory
compared to the other algorithms.
Keywords:
Contextual Information Compression, Context prediction, Vector Autoregressive Model, Principal Component Analysis, Multivariate Vector Model
