Development and Implementation of filtering functions for multiple evaluations of ConCor

Betreuer: Anas Al-Nuaimi
At the LMT we have developed a novel approach for doing robust signal cross-correlation. The algorithm coined ConCor (CONsensus-based cross-CORrelation) has by now been filed at the patent office and published at the ACM Multimedia 2011 conference. We use ConCor to synchronize videos using their bit-rate profiles.

ConCor can be run on multiple filtered versions of the signals to be synchronized. The results obtained indicate the signal feature saliency: if differently filtered versions of the signals always give the same synchronization value with ConCor, it can be expected that the involved signals have salient features and the returned result is trustworthy.

The goal of this work is to evaluate multiple approaches for signal pre-processing of bit rate profiles of mpeg videos to allow for multiple synchronizations and test the synced signals for different properties.

The work will be jointly supervised by Anas Al-Nuaimi, Burak Cizmeci and Florian Schweiger

A contract of 15h/week is desirable
Solid Matlab programming skills
Strong interest in signal processing