Research in BUSIM

A list of research areas and of international projects hosted at BUSIM is given below: 

Biometry based on hand geometry and palmprint; voice based biometry; 2D face biometry;  fusion schemes for multiattribute 3D face biometry; face watermarking.  

Face Processing:
Automatic processing of human faces for man-machine interfaces; face detection in cluttered scenes; facial action coding system and action units; face landmarking; emotion understanding from 2D/3D face images; recognition of head and face gestures; joint gender, age and affect recognition.

Source Separation 
Bayesian source separation for functional optical imaging of brain activities; source separation fort the extraction of cosmic microwave background; source separation in CDMA communication; source separation applications for the detection of defects in textiles and for  earthquake triggering intensity prediction.

3D Shape 
3D Shape descriptors; 3D object indexing and content based retrieval; subspace based methods for shape classification; 3D object retrieval with relevance feedback; 3D object recognition from partial observations.

Semi-fragile watermarking; Text watermarking; watermarking for medical images; steganalysis of audio and image sources;  detection of tampering of information sources; perceptual hashing algorithms; video hashing; blind identification of source devices.

Multiple-Sensor Based Signal Processing
Distributed detection and data fusion for earthquake data.

Medical Image and Information Processing
Tensor Signal Processing, eg. DT-MRI; Medical Imaging and Simulations; Computer Aided Detection and Diagnosis in Radiology; medical IT (eg. PACS systems, teleradiology); volumetric Visualization and User Interfaces.

Sparse Representation and Compressive Sensing

Reconstruction of sparse signal from non-adaptive measurements; Sparse dictionary learning; ell1 minimization

Action Recognition

Human Action recognition from video data


Research Fields

Performance and security improvement with  multibiometrics
Biometry for hand, voice and face; 2D and 3D face processing; multimodality  in biometry; feature extaction and classifier fusion; hash functions 

Automatic processing of human faces for man-machine interfaces
Facial action coding system and Action Units; Conformal mapping; Gabor wavelets; nonlinear registration; face and head gestures


Information Theoretic and Detection Theoretic Analysis of Robust Signal Hashing
We view the robust hashing problem (also known as passive fingerprinting) as a communications problem where the hash value represents the ID of the content to which it belongs. We explore fundamental limits of corresponding communication systems, derive capacity expressions, as well as carry out detection theoretic analysis of a certain class of hash mappings.

Statistical and Information Theoretic Approaches to True Random Number Generation
We focus on a specific class of true random number generator circuits topologies. Relying on certain statistical assumptions on the underlying naturally-occurring continuous-time stochastic process, we analyze and quantify security properties of the resulting bit stream using information theoretic primitives.

Detection Theoretic Aspects of (Active) Signal Fingerprinting
We concentrate on digital fingerprinting in the presence of collusion attacks and model the resulting framework as MAC (multiple access channel). We quantify detection theoretic properties of a number of receivers for various classes of fingerprinting codes.

Robust Digital Watermarking Resistant Against Geometric Attacks
We focus on (improved) spread spectrum watermarking and construct a generalized version of it where the goal is to be able to gain approximate invariance in the presence of translation attacks. We quantify the fundamental trade-offs of the resulting system.

Statistical and Information Theoretic Approaches to Cryptanalysis
We impose a probabilistic setup for a sufficiently broad class of attack approaches. Using asymptotic equipartition property and basic information theoretic primitives, we derive achievable upper bounds on attack complexity.

Robust Hashing of Face Images and Speech Signals
We focus on practical aspects of content authentication. In particular, we focus on face images and speech signals. Via utilizing learning techniques together with meaningful pseudo-random pattern extraction methods, we derive approximately invariant signal features and use them to construct robust hashes.

Content Authentication Using RFID System;
We focus on practical RFID systems, explore limits of existing RFID protocols, as well as propose a number of application scenarios where we use RFID systems to achieve security and authentication.

Applications of Independent Component Analysis in Communications, Image Processing and Geophysics
Content protection applications; Face and lip analysis;  Brain activity classification; Analysis of climatological data

Applications of Source Separation in Image Processing, Communications, and Geophysics
Bayesian source separation for functional optical imaging; Source separation in CDMA communication; Detection of defects in textiles, Restoration of mixed, degraded document images; Earthquake triggering intensity prediction

3D Object Indexing and Content Based Retrieval
Shape descriptors, content-oriented indexation, 3D representations and feature vectors

Design of Steganalyzers for Image and Audio
Perceptual hash functions, watermarking, audio quality metrics, multimedia attacks; audio retrieval

SIMILAR Network of Excellence: Man-machine interfaces similar to human to human communication
Face landmarking; Sign language understanding; 3D Visualization, Interacting with 3D data,; Emotion analysis of speech;

Texture defect detection for textile goods
Texture classification in terms of defective and nondefective, pattern recognition
Modelling of nonstationary and nonGaussian signals
Particle filter modeling, Bayesian modelling, MCMC

Blind Deconvolution
Receiver Function analysis of seismological data, Variational Bayesian approach, sparse signals

Earthquake Early Warning System
Distributed detection under impulsive noise using particle filters and decision fusion using particle filters

Uncertainity reduction in parts-moving problem in robots
Particle filter approach to noise (uncertainity)  reduction for moving robots

Affiliate Groups