 |
Student Projects (Semesterarbeiten)
If you are interested in one of the following topics, please contact the person listed with the respective topic.
If you don't find a compelling topic in our list, we are always keen on hearing about your ideas in the area of our research interests (see the Research section of our website). You are most welcome to discuss your interests with us.
Also, we have a list of finished student projects on our web site.
List of Projects
Comparison of signal separation algorithms (SA)
Signals are often composed of two or more "morphologically" different constituents, e.g., sines and spikes. Various methods for separating these constituents by making use of signal sparsity properties have been described in the literature [1-4].
This project aims at comparing the performance of these sparsity-based methods with classical algorithms used for signal separation, e.g., independent component analysis (ICA) [5], [6]. To this end, one will need to implement the different algorithms and run them on toy, as well as real-world examples such as images corrupted by scratches or fingerprints, mixtures of distinct sounds, or electrical signals emitted by different brain areas.
Type of project: 50% Theory, 50% Simulation
Prerequisites: Good mathematical background and good knowledge of Matlab.
Supervisor: Céline Aubel, Graeme Pope
Professor:
Helmut Bölcskei
References:
[1]
J. Fadili, J.-L. Starck, M. Elad, and D. L. Donoho, "MCALab: Reproducible research in signal and image decomposition and inpainting," Computing in Science & Engineering, vol. 12, no. 1, pp. 44-63, Feb. 2010.
[Link to Document] [2] C. Studer and R. G. Baraniuk, "Stable restoration and separation of approximately sparse signals," Appl. Comput. Harmon. Anal., Jul. 2011, submitted. [Link to Document] [3] G. Kutyniok, "Data separation by sparse representations," in Compressed Sensing: Theory and Applications, Y. C. Eldar, G. Kutyniok, Eds. New York, NY, USA: Cambridge University Press, 2012. [Link to Document] [4] C. Studer, P. Kuppinger, and G. Pope, “Sparse signal recovery from sparsely corrupted measurements,” IEEE Trans. Inf. Theory, May 2012. [Link to Document] [5] A. Hyvärinen and E. Oja, "Independent component analysis: Algorithms and applications," Neural Networks, vol. 13, no. 4-5, pp. 411-430, Jun. 2000. [Link to Document] [6] H. Attias, "EM algorithms for independent component analysis," in Proceedings of the 1998 IEEE Signal Processing Society Workshop , Aug. 1998, pp. 132-141. [Link to Document]
Real-time video surveillance using compressed sensing (SA/DA)
Compressed Sensing (CS) is a way of acquiring sparse signals (e.g. images) by sampling them well below the Nyquist rate. Recent work in [1] has shown how this paradigm can be employed to separate the static part of a video from the moving part. This approach combines classical CS [2] with the problem of completing a low-rank matrix from a small set of its entries [3]. It has the potential to perform significantly better than existing techniques (see [1] and the references therein).
As we want to use these algorithms for a video surveillance system where we wish to separate a moving target of interest from the background, any practical system has to be able to perform the calculations in real-time. Our recent work [4] has demonstrated that this is, indeed, possible.
The goal of this project will be to continue developing the software described in [4], adding more features and ideally porting it to run on Mac OSX (although this last part is not necessary).
Type of project: 25% Theory, 75% Programming
Prerequisites: Good programming skills (experience with Mac OSX programming desirable), familiarity with linear algebra, in particular the Singular Value Decomposition
Supervisor: Graeme Pope
Professor:
Helmut Bölcskei
References:
[1]
E. J. Candès, X. Li, Y. Ma, and J. Wright, "Robust principal component analysis?," arXiv.org.
[Link to Document] [2] E. J. Candès and J. Romberg, "Sparsity and incoherence in compressive sampling," Inverse problems, vol. 23, pp. 969-985, 2007. [Link to Document] [3] E. J. Candès and B. Recht, "Exact matrix completion via convex optimization," Found. of Comput. Math., vol 9, pp. 717-772, 2008. [Link to Document] [4] G. Pope, M. Baumann, C. Studer, and G. Durisi, "Real-time principal component pursuit" 45th Asilomar Conference on Signals, Systems and Computers, 2011. [Link to Document]
Multi-touch touchscreen (SA/DA)
Multi-touch touchscreens allow us to localize multiple touch points simultaneously. The technology has existed since the 1970s but has only reccently become an area of growth with the advent of products such as the iPhone/iPad or the Microsoft Surface [1].
One issue with current
solutions is that they only work well on small surfaces. We recently developed
(in collaboration with the Integrated Systems Laboratory (IIS), ETH Zurich) a light-emitting diode (LED)
based multi-touch touchscreen technology that is scalable to significantly
larger surfaces (i.e., up to several square meters).
Our initial experiments show that multiple touch points can be tracked in a reliable fashion, while
only requiring a small number of LEDs and sensors.
The goal of this project is to build a multi-touch touchscreen prototype based on our method.
To this end, one needs to design, build, and calibrate the LEDs/sensors, attach the LEDs/sensors
to a large LCD, and implement the localization algorithm in Matlab.
If time permits, the localization algorithm can be implemented on an FPGA
for real-time processing.
Type of project: 50% Hardware design (e.g. PCB design and VHDL), 30% Simulation, 20% Theory
Prerequisites: Hardware design knowledge, Matlab, possibly VHDL, and some linear algebra
Supervisor: Michael Lerjen, Graeme Pope
Professor:
Helmut Bölcskei
References:
[1] Microsoft Surface [Link to Document]
Analysis of mass spectrometry data (SA/DA)
In proteomics, researchers use mass spectrometry (MS) measurements to determine which of the possible peptides (and, hence, which proteins) are present in complex sample mixtures.
As there are only very few peptides present in the system at any time point, the data set can be analyzed using sparse signal recovery algorithms.
Previous methods used in the field of proteomics struggle when observing the superposition of several peptides.
In this project, based on the ideas from the rapidly developing area of compressed sensing (CS) [1], sparse signal recovery [2] and dictionary learning algorithms [3], you will develop signal recovery algorithms to identify peptides in MS data.
Such sparsity exploiting algorithms can improve the accuracy of peptide detection significantly and speedup MS data acquisition leading to advances in proteomics.
The project will involve close collaboration with the group of Prof. Aebersold from the institute of Molecular Systems Biology at ETH Zurich.
Type of project: 40% Theory, 60% Simulation
Prerequisites: Good mathematical background required, interest in biological applications of signal processing
Supervisor: Graeme Pope, Veniamin Morgenshtern
Professor:
Helmut Bölcskei, Prof. Ruedi Aebersold (Institute of Molecular Systems Biology)
References:
[1] E. J. Candès and M. B. Wakin, "An introduction to compressive sampling," IEEE Sig. Proc. Magazine, vol. 25, no. 2, pp. 21-30, Mar. 2008. [2] C. Studer, P. Kuppinger, G. Pope, and H. Bölcskei, "Recovery of sparsely corrupted signals," IEEE Trans. Inf. Theory, 2012, to appear . [Link to Document] [3] M. Aharon, M. Elad, and A. Bruckstein, "K-SVD: An algorithm for designing overcomplete dictionaries for sparse representation," IEEE Trans. Sig. Proc., vol 54, no. 11, Nov. 2006. [Link to Document]
Audio de-noising (SA/DA)
Recently, various techniques [1,2,3] based on sparse approximation theory have been proposed for removing specific types of noise and interference from digitized audio signals, such as hum, impulse noise, or quantization noise. At this time, however, it is unclear which of the proposed methods delivers the best quality and how complex the corresponding restoration algorithms are.
In this project, we intend to answer both questions. In the first part of the project, the variety of existing restoration methods should be compared systematically (in terms of de-noising performance). In the second part, you will reduce the computational complexity of the algorithm delivering the best audio quality. In particular, we are interested in the development of an efficient restoration algorithm, which is able to solve the high-dimensional convex optimization problems arising in the problem statement at hand (e.g., using the framework explained in [4]) in real time using an off-the-shelf CPU.
Type of project: Theory and simulation
Prerequisites: Good mathematical background
Supervisor: Graeme Pope, Céline Aubel
Professor:
Helmut Bölcskei
References:
[1] L. Jacques, D. K. Hammond, and J. M. Fadili, "Dequantizing compressed sensing: When oversampling and non-Gaussian constraints combine," IEEE Trans. Inf. Theory, vol. 57, no. 1, pp. 559-571, Jan. 2011. [2] G. Pope, C. Studer, and M. Baes, "Coherence-based recovery guarantees for generalized basis-pursuit de-quantizing," Proc. of IEEE Int. Conf. Acoustics, Speech, and Sig. Proc, Mar. 2012. [Link to Document] [3] A. Zymnis, S. Boyd, and E. Candès, "Compressed Sensing with quantized measurements," IEEE Signal Processing Letters, vol. 17, no. 2, pp. 149-152, Feb. 2010. [4] A. Juditsky, F. K. Karzan, and A. Nemirovski, "l1 Minimization via randomized first orders," Optimization Online, 2010. [Link to Document]
Implementation of a wireless relaying system (SA/DA)
The Communication Theory Group is developing a hardware prototype platform for a wireless relaying system employing multiple-input multiple-output (MIMO) and wireless local area network (WLAN) technology. Various student projects are possible around this platform which allows you to gain hands-on experience with the latest wireless technologies and research results.
The main goal of the overall platform project is to implement elements of a wireless relaying infrastructure. Depending on your skills and interests,
your project can be adapted and can include RF and hardware development, VHDL implementation, or signal processing with Matlab.
The relay nodes are built from dedicated boards that contain an RF chain and an FPGA for baseband signal processing. In a first step, the platform will be set up with several RF nodes and signal processing will be done offline on a PC in Matlab.
Type of project: Development of hardware (RF, FPGAs), firmware (VHDL), or software (Matlab) is possible
Prerequisites: Interest in wireless communication systems
Supervisor: Michael Lerjen
Professor:
Helmut Bölcskei
References:
[1] V. I. Morgenshtern and H. Bölcskei, "Crystallization in large wireless networks," IEEE Trans. Inf. Theory, vol. 53, no. 10, pp. 3319-3349, Oct. 2007. [Link to Document] [2]
C. Akçaba, P. Kuppinger, and H. Bölcskei, "Distributed transmit diversity in relay networks," in Proc. of IEEE Information Theory Workshop (ITW), Bergen, Norway, pp. 233-237, July 2007. [Link to Document]
|
|
