SGN-1706 SIGNAL PROCESSING PROJECT, 5-8 CR

2011-2012

Project Topics

Using Microsoft Kinect SDK for Gesture Recognition
Simulation of FIR Filter Structures
Measuring and Modeling Gene Expression in Individual Bacteria
GPU-prosessointi
Testing humans allergies from digital skin images
Computational Systems Biology
Sound direction of arrival (DOA) algorithm
Deep sequencing data compression
Signal processing for deep sequencing data analysis
Bio-inspired optimization algorithms
Implementation of Flux Coupling Finder (FCF) framework

Image content classification using segmentation and shape description This topic is reserved by Ahti Oksanen.
Integrative analysis of prostate cancer sequencing data This topic is reserved by Janne Seppälä.
Measuring Gene Expression in Individual Bacteria This topic is reserved by Adrien Sala.
Analysis of Divergent Promoters at the Single Molecule Level This topic is reserved by Anisha Viswanathan.
Adaptive methods for contrast enhancement of low quality videos This topic is reserved by Weiyi Xie.

Using Microsoft Kinect SDK for Gesture Recognition

Background:
Microsoft Kinect is a low cost gaming device originally developed for MS Xbox 360. Soon after its release independent developers hacked the protocol and released hack drivers for PC. The driver package was soon adopted for many different projects and demos (see, e.g., Youtube: 12 BEST Kinect HACKS).

Partly because of the hack drivers, in June 2011 Microsoft released their official MS Kinect SDK. While the hack drivers included only raw streams (VGA and depth map images), the SDK features also skeletal tracking, which fits a skeleton into all human figures whenever they become visible for the device (see an example of skeleton tracking result). This increases the level of abstraction and allows for new kinds of applications.

Objectives:
In this project, the task is to familiarize yourself with the Kinect SDK, and implement a simple gesture classifier on top of it. In other words, you will prepare training data of a selected gesture (e.g. pointing your hand to the right) and training data of other gestures. Using the data, prepare a classifier that detects the gesture and acts accordingly. The feedback in this demo can be e.g., tilting the device upwards or downwards (the device has a tilt motor that can be controlled via the SDK).

Alternatively, if you come up with a more innovative Kinect SDK demo, please feel free to propose.

The device will be provided by the client.

Group size:
Approx. 2 persons. Requires C++ programming and pattern recognition skills.

Client:
Heikki Huttunen
Department of Signal Processing, TUT
Room TE408
firstname.lastname at tut dot fi

Simulation of FIR Filter Structures

Background:
Demand of low power consumption in e.g., electronical devises leads to a demand of efficient filter structures and their design techniques, especially in applications demanding sharp FIR filters.

Efficient VLSI circuits for digital filters can be generated by using structures requiring few general multipliers which leads to smaller silicon area, lower power consumption and higher sampling rate.

For digital filters, the major power consumption is due to arithmetic operations such as multipliers and adders or subtractors. Computationally efficient recursive digital filter structures have been proposed for the implementation of FIR filters.  The structures are more suitable for dedicated HW than DSP processor implementation. The aim is to simulate such filter structures regarding to finite-word length effects.

Objectives:

Simulation guidelines and structures are given by the client.

Group size:
1-2 person

Required deliverables:
In general, deliverables should consists of

Prerequisites:
Fair knowledge of Matlab and digital filters.

Client:
Raija Lehto
Department of Signal Processing, TUT
firstname.lastname at tut dot fi
TE421

Measuring and Modeling Gene Expression in Individual Bacteria

The Biosystem Dynamics Laboratory focuses on the study gene expression dynamics via measurement and modeling. The student will be involved in the measurement of gene expression at the single bacteria level, modeling the dynamics with detailed models at the single molecule level, and understanding the regulatory mechanisms of gene expression in E. coli. Depending on the student’s background the work will focus more on his/hers area of expertise and personal interests. This work can be both theoretical and at our microbiology laboratory. A successful project may
lead to an opportunity to do summer internship or diploma thesis work at the CSB research group.

Client:
Andre S. Ribeiro
Department of Signal Processing, TUT
firstname.lastname at tut dot fi
Room TC336

Analysis of Divergent Promoters at the Single Molecule Level

Background:
The kinetics of transcriptional events depend on gene sequences, structure of the DNA strand, strength of repression, as well as the concentration of metabolites in the cellular environment. Characterizing the molecular dynamics of divergent promoters in E.coli, using single cell molecule detection methods in transcription will allow better understanding of how metabolite affects the transcriptional dynamics in-Vivo. We will investigate, at the single cell, single promoter level, the behavior of divergent promoters in E. coli, and the consequent relative expression levels of different genes. This
 will involve either measurement of gene expression at the single bacteria level, image analysis and data analysis, or modeling, depending on the student’s background and will aid in understanding the dynamics of gene expression in E. coli.

Objectives:
1. Literature review, executing the cell culturing and microscopy experiments. (References also available from the client)
2. Analysis of microcopy images and interpretation of the data (Necessary support will be provide by the client)

Group size: 2 persons
Person 1: Basic knowledge in microbial cell culture technique experiences and microcopy.
Person 2: Fair programming skills in MATLAB.

Clients:
Meenakshisundaram Kandhavelu and Andre S. Ribeiro
Department of Signal Processing, TUT
firstname.lastname at tut dot fi
Room TC336

GPU-prosessointi

Tausta:
Modernit pc-tietokoneet ovat signaalinkäsittelyssä ja tietoliikenteessä yllättävän suorituskykyisiä, mutta tietyissä yksinkertaisissa tehtävissä suorituskyky loppuu. Olemme kiinnostuneet tutkimaan erittäin korkean näytteenottotaajuuden signaalien käsittelyn siirtämistä grafiikkaprosessorille (GPU). Lisätietoa tekniikasta esimerkiksi osoitteesta: www.nvidia.com/page/hpc.html.

Tavoitteet:
1) Toteuttaa valitut tehtävät GPU:lla.
2) Selvittää GPU:n suorituskyvyn/ohjelmoitavuuden rajat kyseisissä tehtävissä.

Ryhmän koko:
1-2 opiskelijaa.

Tuotokset:
1) Ohjelmakoodi (C/C++)
2) Demonstraattori
3) Raportti

Esitiedot:
Ohjelmointikokemus pakollista, tietoliikennetekniikan tunteminen suositeltavaa.

Muuta:
Tilaaja toimittaa testimateriaalin ja soveltuvan näytönohjaimen.Tekijän toivomme omistavan PC:n vapaalla PCI-E -korttipaikalla. Menestyksekkäästä työstä mahdollisuus kompensaatioon.

Yhteystiedot:
Janne Mansikkamäki, Insta DefSec Oy
etunimi.sukunimi@insta.fi

Adaptive methods for contrast enhancement of low quality videos

Background:
Low contrast is a typical problem with videos recorded in poor illumination conditions. However, the visibility of even the dimmest details is highly important in application areas such as forensics or biomedical imaging. Contrast enhancement aims to increase the visibility of details that may be obscured by deficient global or local illumination.

Objectives:
The group will familiarize themselves with the state-of-the-art global and local contrast enhancement techniques, and select few of them (adjusted based on the group size). Thereafter, the selected algorithms will be implemented and their performance will be studied. The implementation will be done using the GStreamer open source multimedia framework (http://www.gstreamer.net).

Group size:
1-2 persons. Workload will be adjusted according to group size. Requires C programming skills (preferably in Linux environment).

Clients:
Pekka Ruusuvuori
Department of Signal Processing, TUT
firstname.lastname at tut dot fi

Antti Lehmussola
Forensic laboratory, National bureau of investigation

Testing humans allergies from digital skin images

Brief description:
Various allergies are very common among people in industrialized countries. In order to diagnose, research and treat them, clinical tests have to be performed. The most common of these tests is the skin prick test where the allergen is placed on the skin and the skin is pricked by a needle. The size of the possibly emerging skin reaction is then measured.

A method to measure the skin reaction size has been developed at TUT. The method is based on photographing the skin area and analyzing the
photographs.

The aim of this project is to find a good and reproducible photographing setup (camera, objective, lighting) for this purpose.

In order to validate the setup, the project group has the opportunity to test the setup at the Skin and allergy hospital in Helsinki. The images thus acquired will be analyzed with the algorithm developed earlier.

A basic understanding of photography and basic Matlab skills help in the project.

Group size:
2-3 students

Client:
Heikki Forsvik
Department of Signal Processing, TUT
firstname.lastname at tut dot fi

Computational Systems Biology

Computational systems biology research group offers a wide variety of project work topics. Topic can be individually specified to correspond to the interests of the student. Possible topics include (but are not limited to) projects from biological image processing, simulation and modeling of biological systems, analysis of biological e.g. microarray data, and developing different kind of software tools. Tasks can be defined to be mathematical or explanatory in nature, depending on individual interests. Some projects can be done in collaboration with companies.

Objectives:
1) To be specified individually based on the topic of the project work.

Group size:
1-3 persons

Required deliverables:
Typical deliveries include Matlab or R implementation of the solution and a written report.

Prerequisites:
Varies, typically at least basic skills in Matlab programming are required. Student should have an interest in studying the biological systems and the course SGN-6106 Computational Systems Biology is recommended.

To get more information on individual topics contact:
Olli Yli-Harja
Department of Signal Processing, TUT
firstname.lastname at tut dot fi
Room TE306

Sound direction of arrival (DOA) algorithm

Background:
Sound field is a vector field, represented by sound pressure and particle velocity. In noise control applications the sound direction of arrival (DOA) is often of interest. Traditionally, the DOA is estimated using a pressure microphone array. More recently the particle velocity approach has gained popularity. Particle velocity and sound pressure are related with Euler's equation of motion. Your task is to investige the use of closely spaced pressured microphones to estimate the DOA via particle velocity.

Objectives:
1) Literate survey on the subject
2) Collect data in the audio laboratory using a Soundfield microphone
3) Implement a selected method in Matlab
4) Analyse method's performance

Group size:
1 person

Required deliverables:
1) MATLAB codes of algorithm
2) Report

Prerequisites:
Fair skills in MATLAB programming

Client:
Pasi Pertilä
Department of Signal Processing, TUT
firstname.lastname at tut dot fi
040 8490 786

Deep sequencing data compression

Background:
Modern biological measurement systems produce massive amounts of data,
typically in the form of short DNA sequences, e.g.:
     ATGCGATGATCGAATGGGATCGAGTGGGAT
     CAGTGAGCGGGGCCCTAGAACGCCCAAGAT
     AGGTTATCCTATATTTGTTACATATTTANC

Output file of a single experiment can be several hundred gigabytes and often multiple experiments are run in parallel. The storage of such amounts of data is a major problem. With general purpose compression tools, such as gzip, compression ratio of 1/4 is typically obtained. However, several approaches that utilize the redundancy in the data have been developed and compression ratios up to 1/300 have been reported.

Objectives:
1) Literature review on available tools and methods for deep seqencing data storage, focus on algorithms that are basis of the tools
2) Benchmark different implementations in research environment
3) Proof of concept level setup of such data storage system in reseach
environment, that allows to deposit and loading of data.

Prerequisites:
Knowledge of basic unix tools than are used to work with large text files

Group size:
1-3 person

Client:
Matti Nykter
Department of Signal Processing, TUT
firstname.lastname at tut dot fi

Signal processing for deep sequencing data analysis

Backgroud:
Modern biological measurement systems produce massive amounts of data,
typically in the form of short DNA sequences, e.g.:
     ATGCGATGATCGAATGGGATCGAGTGGGAT
     CAGTGAGCGGGGCCCTAGAACGCCCAAGAT
     AGGTTATCCTATATTTGTTACATATTTANC

Analysis of such data includes several non-trivial steps that can be addressed using established signal processing methodologies. First step in the analysis is to uncover the origin of the above type of sequences, or "reads". This can be done by comparing reads to reference genome. For human the size of reference genome is about 3 billion letters. Thus, comparison is non trivial. Established solutions are based on hashmap or burrows-wheeler transformation. Subsequent analysis includes quantication of the number of reads from a specific area of genome and detecting the areas with more reads than expected by change. This can be done, for example, with methods such as signal/peak detection, classification, or regression models.

This project will focus on a single step of this analysis pipeline, and project group should provide state of the art solution for a specific real life data analysis problem. Solution can be an implementation based on published methods or authors can propose a novel solution that should be benchmarked against standard tools.

Objectives:
1) Literature review of methods proposed for specific data analysis problem
2) Implementation of the method of choise
3) Bechmarking with real data

Prerequisites:
Project does not require special knowledge from biological domain. Client will provide necessary data in readily accessable format. Knowledge on suitable signal processing methods and fair skills in matlab programming are required.

Group size:
1-3 person

Client:
Matti Nykter
Department of Signal Processing, TUT
firstname.lastname at tut dot fi

Bio-inspired optimization algorithms

Overview:
During the last two decades, biology has played an important role in computational sciences, through the impact that biologically inspired algorithms had on the optimization techniques. Metaheuristics such as ant colony optimization, evolutionary algorithms, particle swarm optimization, bee colony optimization have been developed and applied to complex system modelling. Starting with classical optimization problems such as the traveling salesman problem and arriving at real time optimization of the airport boarding process these algorithms have shown their capabilities. However, their applicability to divers problems needs to be assessed and carefully tested.

The current topic consists of comparing several bio-inspired optimization algorithms for image matching.
Image matching is widely applied in the areas of pattern recognition, computer vision, medicine, remote sensing, aircraft navigation and movement tracking.

Objectives:
1) Literature survey of the topic and given algorithms
2) Matlab/C-implementation of the algorithms
3) Evaluation of the algorithms in Matlab/C

Prerequisites:
Fair skills in Matlab and C programming.

Client:
Adriana Vasilache, Nokia Research Center
firstname.lastname at nokia dot com

Implementation of Flux Coupling Finder (FCF) framework

Background:
The recent abundance of complete genome sequences has enabled the production of reconstructions of metabolic networks for various organisms. Examination of the structural and topological properties of metabolic networks is important at both the conceptual level, to reveal the organizational principles of metabolic interactions within cellular networks, and at the practical level for more effectively focusing engineering interventions and ensuring the consistency of the underlying reconstructions.

Flux coupling finder is able to determine metabolic fluxes which are coupled together and which are completely blocked. This functionality is useful in various applications, e.g., in examining the correctness of metabolic reconstructions, and in identifying genetic knockouts which have equivalent effects in metabolism. The approach uses linear programming optimization and the work is implemented using Matlab.

Article describing the method: http://genome.cshlp.org/cgi/content/abstract/14/2/301

Group size:
1-2 persons

Client:
Tommi Aho
Department of Signal Processing, TUT
Room TF412
firstname.lastname at tut dot fi

Image content classification using segmentation and shape description

This topic is reserved by Ahti Oksanen.

 Background:
The one-dimensional function which is derived from shape boundary coordinates usually captures the perceptual feature of the shape [2]. In this project we segment the region of interest of the image and then use the boundary signature to describe and classify the objects. Additionally, a set of topological descriptors can be used to improve classification performance. (See example 1 and example 2.)

Objectives:
1) Literature survey on the subject (at least reference [2]).
2) Collecting imag data in the laboratory TE407 using Basler a201bc camera.
3) Implementing the set of tasks using Matlab.

(The details will be defined by 12.9.2011)

Group size:
2 persons

Required deliverables:
1) One Matlab program = GUI with all components (not compiled) such that the reference platform is the current Matlab version in TC303.
2) One report that includes short documentation for each code component and user instructions for the GUI.
In addition to text, the report contains pictures and flow charts that explain each processing step in detail and pictorially.

Prerequisites:
1) Image processing skills in MATLAB programming. A set of useful scripts will be provided by the client.
2) GUI-programming skills are beneficial, but can be learned during the project, if necessary.

Client:
Petri Hirvonen
Teaching associate
Department of Signal Processing, TUT
Room TF308
firstname.lastname at tut dot fi

References:
[1] = http://en.wikipedia.org/wiki/Microscope_image_processing

"Microscope image processing is a broad term that covers the use of digital image processing techniques to process, analyze and present images obtained from a microscope."

[2] = http://hal.archives-ouvertes.fr/docs/00/44/60/37/PDF/ARS-Journal-SurveyPatternRecognition.pdf


Main page