Site Sections:  oWelcome  oDepartment  oStudies  xResearch  oPublications  oNews  oSearch  ( [FI] Suomeksi )

Shortcuts:

Research Areas

Signal processing in general has greatly changed during the last fifteen years and it will change even more in the coming years. What was earlier viewed as digital signal processing now forms only a small part of the new concept of signal processing which might be more adequately described as the methods of analyzing, manipulating and presenting natural information. The new communication and imaging technologies have vastly increased the capabilities of individuals to interact with the surrounding world in a multimodal way, and the recent advances in biology, such as microarrays, have opened ways to a true understanding of biological processes. The future of signal processing is in developing techniques that enable natural interaction using modalities that are most natural for a human, and in developing modelling tools for biological systems. The key to success, however, is in combining new visions with strong theoretical foundations and understanding the needs of practical implementations.

The Department's research areas are dynamically changing and modifying in order to meet these new challenges.

Speech and Audio

Research on speech and audio signals includes speech, music, and spatial processing. Concrete research topics include for example Finnish speech synthesis and voice modification, automatic music transcription, and speaker localization using multiple microphones. The research is carried out by Audio Research Group.

More on the group's research can be found on their research page.

Spectral and Algebraic Techniques in DSP

Spectral and algebraic methods play a central role and mark a large and important field in signal processing. Broadly speaking, they comprise any harmonic analysis applied to signals of interest. We consider the synergy of abstract mathematical constructions, such as function space bases, frames, function decompositions and projections, with their practical implementations through fast discrete structures.

Particular topics of interest are listed below

  • Efficient Signal Representations
    • Spline-based methods for signal and image processing
    • Fast algorithms for signal and image processing
    • Adaptive nonparametric techniques for varying scale image and signal processing
    • Algebraic framework for hyper-harmonic analysis of signals and systems
    • Fibonacci transforms, decision diagrams and interconnection topologies
    • Spectral transforms and logic design
  • Transform Methods for biomedical signal processing
    • Signal processing methods for brain data
    • ECG denoising, compression and micro-potentials recovery
  • Matrix algebra methods in signal processing
    • Combinatorial methods in signal processing
    • Linear least square problems by partial orthogonalization process
    • Construction and decomposition of orthogonal arrays and Hadamard matrices
  • Efficient methods for signal and image de-noising and enhancement
    • Transform domain adaptive filtering
    • Efficient fractal-based image coding algorithms
    • Wavelet and wavelet-packet image coding algorithms
    • Efficient encoding of wavelet coefficients significance maps
    • Wavelet transform method for coding images corrupted by film-grain noise

The research is conducted by Transforms and Spectral Techniques Group.

Nonlinear Signal Processing

Nonlinear filtering techniques based on polynomial and order statistics continue to be a major research topic of the department since many years ago. Recently, the focus has been on rational filtering and some hybrid structures involving rational and order statistics (OS) subfilters. Ad-hoc structures have so far been investigated and the results in image processing are promising. The following topics further explain the work done. Future work in this field will focus on further analysis of such structures with the aim of developing optimization tools for given tasks in image processing.

  • Adaptive fuzzy scalar median-hybrid filters for grey level image restoration
  • Design and real-time implementation of recursive and non-recursive 3-D Median-rational hybrid filters
  • Fuzzy OS-Rational filter for multichannel image restoration
  • Nonlinear unsharp masking using rational control function for noisy color image enhancement
  • Programmable hardware system for OS-Rational filters using RNS and DCVS logic.

The research is conducted by the Image and Video Analysis Group.

Video Analysis and Coding

The video team has been working on the following topics:

  • Error concealment
  • Sub-picture video coding for unequal error protection
  • Coding of faded scene transitions
  • Picture buffering for prediction references and display
  • Sub-sequences for video coding
  • Isolated regions in video coding
  • Study on H.264/AVC performance in DVB-H IP data casting service
  • Study on H.264/AVC performance in 3GPP Multi/Broadcast service (MBMS)

The research is carrying out by the Image and Video Analysis Group.

Content-based Indexing and Retrieval over Large Image and Video Databases

Interactive multimedia services will strongly influence and even dominate the future of communications and telecommunications. Both the flexibility and efficiency of the coding systems used, as well as the ability to efficiently search for particular content of interest on distributed data-bases are essential for the success of these emerging services.

Since the early 1990s, content-based retrieval of digital imagery has become a very active area of research.

In this context we have been developing the MUVIS, which is a system for content indexing and retrieval for mutlimedia databases.

Group involved: The Image and Video Analysis Group.

Compression Techniques

  • Source compression
    • Structured lattice vector quantization
    • Computation and quantization of short term spectral parameters of speech
    • VQ image coding and index assignment over memory channels
    • Lossless signal compression (audio, ECG, images)
    • Variable-to-fixed coding: asymptotic theory and efficient implementation
    • Modeling with distortion
  • Estimation, Optimization and Adaptive filtering
    • Convex optimization in signal processing
    • Optimizing the coding gain of the filter-bank for a given input spectrum
    • Semi-definite programming based optimal filter-banks with infinite impulse response

Statistical Modeling and Inference Group.

Statistical Inference for Genomics

  • Normalized ML (NML) models for discrete regression
  • Classification and prediction with gene expression data
  • Genomic network modeling and gene clustering from time series microarray data
  • Biological sequence compression
  • Genomic data analysis

The Statistical Modeling and Inference Group

Interactive Multimedia Technology

Interactive multimedia is a quickly developing technology, which gives to the user access to digital content via a network. This technology integrates signal processing with telecommunications and software engineering. It requires a broad knowledge of new developments in these disciplines and mastering the use of practical software and hardware tools. The Department of Signal Processing has excellent laboratory facilities for this kind of research including media server, broadband ATM network, high-end workstations, video compression and decompression systems.

The recently built Digital Media Laboratory has been continuously expanding. The laboratory has state-of-the-art equipment for digital television and multimedia networking research. A tested system for digital television distribution in IP networks has been constructed. The system uses local area network of the Department of Information Technology to test the distribution of a large number of digital television channels. Television multiplexes are received from satellite and terrestrial broadcast by distribution servers. There are six servers, each with five receiver cards. The servers extract transport streams of TV programs and send them over the network in IP multicast. The tests indicate that the system performs well with about 50 television streams broadcasted to many users. The Digital Media Laboratory has been also equipped with a system for generating digital television service based on the Multimedia Home Platform (MHP) standard. This system is composed of a MHP application generator, a Transport Stream generator and a QPSK modulator making it possible to watch MHP applications on a set-top box. The presentation room of Digital Media Laboratory has high-resolution plasma displays and an LCD projector. The system is fully integrated and it is operational.

The Multimedia Group.

Multimedia and Data Mining

The Multimedia and Data Mining Group.

Health Informatics

Major research topics:

  • Analysis of physiological signals during sleep
  • Analysis of physiological and other signals from new sensors
  • Medical device communication standardisation

The Sleep and Sensory Signal Analysis Group.

Methods and Models for Biological Signals and Images

Research in this topic concentrates mainly on medical image reconstruction, processing and analysis. The aim is to develop automatic methods to analyze 3-D functional images. The research is carried out by:

The Methods and Models for Biological Signals and Images Group.

Design and Implementation of Digital Filters and Filter Banks for Various Applications

Various optimization techniques for filters and filter banks are studied thoroughly in order to achieve efficient and effective filter structures

The Digital Filtering Group.

Virtual reality, perceptual user interfaces and 3D graphics

Research Milestones:

  • A running web system integrating service databases, 2D maps, and 3D models of a city.
  • One of the first mobile 3D navigation systems for city environments.
  • An efficient 3D terrain rendering engine and an editing tool for 3D scenes and events.
  • A novel high-quality walk-thru fog screen.

The Virtual Reality Group.

Signal Processing for Systems Biology

Research in this topic consists of applications of signal processing, image analysis and various computational and machine learning methods for the systems biology field, most importantly in functional genomics and cell signaling. We develop modeling and simulation tools for studying complex biological phenomena. We also apply the methods in modern biological measurement techniques, e.g. gene and cell array technologies and different types of microscopy.

The systematic modeling approach involves all aspects of the problem domain: The underlying biological processes, modern measurement systems, efficient signal processing tools, analysis and interpretation of the data, and advanced mathematical models. Our strength lies in understanding all of these processes and being able to combine them meaningfully. The research team combines experts from different fields - biology, medicine, bioinformatics, signal and image processing, statistics, mathematics, and computer science.

Research topics:

  • Dynamic models of genetic signaling networks
  • Modeling tools for describing the new measurement techniques
  • Signal and image processing tools for emerging techniques

The research is conducted by:

The Computational Systems Biology Group.