Agenda

CAS MSc Midterm Presentations

Max Schöpe

Additional information ...


Signal Processing Seminar

Signal processing in distributed networks; audio signal processing

Metin Çalış

Additional information ...


Signal Processing Seminar

Biomedical signal processing

Aybüke Erol

Additional information ...


CAS MSc Midterm Presentations

Tracking of rodents from infrared video sequences for behaviour studies

Javier Guinea Perez


Signal Processing Seminar

Developing a CT Scanner Detector Panel Based on Photon Counting

Dennis Schaart
Applied Physics, TU Delft

Additional information ...


MSc SS Thesis Presentation

Distributed Coordination for Multi-feet Truck Platooning

Yikai Zeng

Truck platooning refers to coordinating a group of heavy-duty vehicles at a close inter-vehicle distance to reduce overall fuel consumption. This coordination between trucks is traditionally achieved by adjusting the schedule, velocity and routines to increase the platooning chances, and thus improve the overall fuel efficiency. However, the data model built for the coordination problem is typically integer-constrained, making it generally hard to solve. On the other hand, the interaction among self-interested fleets which are operated by different companies is not well-studied. This thesis aims to build a distributed framework for multi-fleet truck platooning coordination to enable the coordination without a third-party service provider.

The interaction among fleets is considered a non-cooperative finite game, for which we propose the best response search method, which essentially requires to solve a cooperative truck platooning optimization problem iteratively. We refer to the optimization problem as a best-response problem, which is formulated as a mixed-integer linear problem with relaxation skills.

 To achieve a feasible time complexity for the best-response subproblem, we propose a decentralized algorithm, distributing the computational load to connected automated vehicles within the fleet.

The proposed method is examined under a real-world featured demand set to compare the performance in optimality and time complexity with previous studies. The result suggests that the decentralized algorithm delivers the optimal objective value in this case, while the best-response search does not deliver extra benefits as a the dominating time costs in the cost functions eliminate the potential for improvement.


MSc SS Thesis Presentation

Atrial Fibrillation classification from a short single lead ECG recordin

Yuchen Yin

This thesis focuses on classifying AF and Normal rhythm ECG recordings. AF is a common arrhythmia occurring in millions of people every year, which could lead to blood clots, stroke or even heart failure. When AF is occurring, the P waves are often absent and RR intervals are often irregular.

This thesis proposes a new Poincaré plot based feature that exploits the distribution and position information of the plot. The Poincaré plot can visually analyze the nonlinear aspects of the heart rate dynamics both qualitatively and quantitatively. In this thesis, the Poincaré plot values are first quantized into small bins, which represent whether corresponding states are visited by the system or not, by setting ones or zeros. The bins are then given weights by the masks based on the probability of each state being visited by the system, and the relative position between the bins and the center of the plot. By calculating the element-wise multiplication and summation between the quantized Poincaré plot and the masks, the expected value of the matrix of the quantized Poincaré plot is computed, and the outliers in the plot are emphasized. Therefore, the proposed feature is assumed to have a higher value for the AF rhythms and a lower value for the Normal rhythm.

  Instead of RR intervals, the Poincaré plot used in this thesis is also generated from the peak intervals in the autocorrelation function of both ECG and prediction error. The autocorrelation function aims to evaluate the self-similarity of the ECG signals and thus extracts the irregularity of the AF signals.  

The dataset used in this thesis comes from the Physionet Challenge 2017, containing 5076 Normal recordings and 758 AF recordings. In total, 21 Poincaré plot based features are used to train the SVM and random forest models, which yields the F1 score of 0.80 and 0.85, respectively. When using features from the same intervals, RR intervals generate the highest F1 score of 0.77 and 0.81, followed by the peak intervals in the autocorrelation of prediction error with the F1 score of 0.74 and 0.78, followed by the peak intervals in the autocorrelation error of ECG with the F1 score of 0.63 and 0.68. Using the minimum redundancy maximum relevance algorithm, eleven features are selected based on their importance. Training the SVM and RF models with these features reaches the F1 score of 0.78 and 0.84, respectively.

Additional information ...


MSc SS Thesis Presentation

Clock skew invariant beamforming

Laurens Buijs

This thesis is focused on Wireless Acoustic Sensor Networks (WASNs) used for beamforming in a speech enhancement task. Since each node in aWASN has its own clock, clock osets and clock skews between the nodes are inevitable. Clock osets and clock skew can be detrimental to the beamformer performance. In this thesis we focus on the eect of clock skew on the beamformer performance. Existing methods for clock skew compensation for the speech enhancement application do this explicitly. In this thesis we investigate the possibility to formulate the beamformer such that explicit clock skew compensation is not necessary.

Instead, we propose an algorithm for implicit clock skew compensation, which takes advantage of the Generalized Eigenvalue Decomposition (GEVD) to construct beamformers (e.g. Minimum Variance Distortionless Response (MVDR)), recently proposed in the literature. Using the GEVD, no explicit compensation has to be applied to the received data. Compared to the state-of-the-art, where clock skew estimation/compensation algorithms are used, this reduces the computational complexity for beamformer processing.

The algorithm depends on exact knowledge of the noisy correlation matrix across the microphones. In practice, this matrix is unknown and estimation will reduce the performance of the proposed algorithm. We therefore quantify the error made in the estimation of the correlation matrix using the standard Welch method and also look at a recursive smoothing based method for correlation matrix estimation. Compared to a selected state-of-the-art algorithm, the proposed algorithm shows similar or better performance using this recursive smoothing method. For future work on this subject, more study can be done on correlation matrix estimation methods, as these play a key role in clock skew invariant beamforming.

Additional information ...


Signal Processing Seminar

Biomedial Signal Processing

Hanie Moghaddasi

Additional information ...


MSc SS Thesis Presentation

Ultra fast MRI acquisition at 7 Tesla

Alejandro Monreal Madrigal

Magnetic Resonance Imaging is one of the most widely used imaging modalities nowadays and it performs especially well imaging human organs such as the brain and liver. One of its main limitations is the relatively long imaging times, to overcome this issue and speed up the data acquisition, several techniques such as Parallel Imaging or PI have been developed. These techniques require advanced hardware and software to be able to decrease the acquisition time. On the hardware side, a highly efficient insert gradient coil has been designed and built at the University Medical Center Utrecht. Specialized software has to be implemented to optimally make use of this hardware. One of the recently proposed PI methods called Wave-CAIPI has been proved to achieve a ninth fold acceleration factor without compromising image quality.  

This project aims to investigate the time gain that can be achieved when combing the insert gradient coil with a Wave-CAIPI strategy. Two main aspects are reviewed. The first one is the maximum achievable under-sampling factor that does not compromise image quality. The second one is the decrease in acquisition time that can be obtained when using the insert gradient coil compared to conventional gradient systems while maintaining image quality. To do so, the strategy has been implemented and extensive simulations have been performed to optimize the MR acquisition parameters. To prove the results from the simulations, the Wave-CAIPI sequence was implemented in a 7T scanner at the UMCU, where the acquired data was retrospectively under-sampled, obtaining the wave image to be further reconstructed.


MSc SS Thesis Presentation

Automatic Depth Matching for Petrophysical Borehole Logs

Aitor García Manso

In the oil and gas industry a crucial step for detecting and developing natural resources is to drill wells and measure miscellaneous properties along the well depth.  One important source of this disturbances is depth misalignment and in order to compare different  measurements care must be taken to ensure that all measurements (log curves) are properly positioned in depth. This process is called depth matching. In spite of multiple attempts for automating this process it is still mostly done manually.   

Based on the Parametric Time Warping (PTW), a parameterised warping function that warps one of the curves  is assumed and its parameters are determined by solving an optimization problem maximizing the cross-correlation between the two curves. The warping function is assumed to have the parametric form of a piecewise linear function in order to accommodate the linear shifts that take place during the measurement process. This method, combined with preprocessing techniques such as an offset correction and low pass filtering, gives a robust solution and can correctly align the most commonly accruing examples. Furthermore, the methodology is extended to depth match logs with severe distortion by applying the technique in an iterative fashion. Several examples are given when developed algorithm is tested on real log data supplemented with the analysis of the computational complexity this method has and the scalability to larger data sets.


MSc SS Thesis Presentation

Optimal Sensor Placement for Calibration-Involved Radio Astronomy Imaging Applications

Kaiwen Zhang

In radio astronomy (RA), one of the key tasks is the estimation of the celestial source powers, i.e. imaging. To maximize the performance, it is crucial to optimize the receiver locations before the construction of a telescope array. However, although system calibration is an integral and crucial process of imaging, it has rarely been addressed for RA sensor placement problems previously. This motivates us to investigate whether incorporating calibration can result in better array designs.

In this thesis, we focus on the calibration of the sensors’ complex-scalar gains in particular, which are treated as nuisance parameters for the image estimation. The associated Cramer-Rao bound (CRB) is derived and employed as the design criterion. The nonlinear CRB-based sensor placement problem is cast as an NP-hard combinatorial optimization problem, and we adopt two approaches to solve such by approximation: (i) greedy submodular maximization and (ii) convex optimization with semidefinite relaxation. The former is chosen for simulations due to its good performance and lower computational complexity. Extensive simulations demonstrate that compared to the calibration-excluded design, the proposed one only provides slight improvements to the imaging quality. However, the proposed array demonstrates the potential of accelerating the convergence of the gain estimation procedures. Through further investigation, we conclude that the lack of imaging quality improvident can be a consequence of the gain and image being near-orthogonal parameters.


MSc SS Thesis Presentation

Low-field MR Imaging Using a Nonuniform Fast Fourier Transform

Maria Macarulla Rodriguez

Low-field Magnetic Resonance Imaging (LF MRI) is a cheap and safe technique to visualise the internal structure of the human body. Unlike other imaging techniques, Magnetic Resonance Imaging does not use ionising radiation to generate the images. Instead, it uses magnetic fields and radio waves which are nonthreatening to the health. The LF MRI scanners are constructed out of inexpensive materials and their maintenance is affordable. Therefore, these scanners are a promising alternative for developing countries that present economic limitations. Nonetheless, since Magnetic Resonance scanners use a weak magnetic field, the process of image reconstruction requires complex algorithms that need time. This thesis will examine the way in which the computational time of the image reconstruction from a low-field Magnetic Resonance Imaging can be reduced, using an algorithm based on the fast Fourier transform.


Signal Processing Seminar

Second Master presentation

Maria Macarulla Rodriguez, Aitor García Manso