Agenda

PhD Thesis Defence

Violeta Prodanovic

Additional information ...


PhD Thesis Defence

Pan Liu


PhD Thesis Defence

Multi-Microphone Noise Reduction for Hearing Assistive Devices

Andreas Koutrouvelis

The paramount importance of good hearing in everyday life has driven an exploration into the improvement of hearing capabilities of (hearing impaired) people in acoustic challenging situations using hearing assistive devices (HADs). HADs are small portable devices, which primarily aim at improving the intelligibility of an acoustic source that has drawn the attention of the HAD user. One of the most important steps to achieve this is via filtering the sound recorded using the HAD microphones, such that ideally all unwanted acoustic sources in the acoustic scene are suppressed, while the target source is maintained undistorted. Modern HAD systems often consist of two collaborative (typically wirelessly connected) HADs, each placed on a different ear. These HAD systems are commonly referred to as binaural HAD systems. The noise reduction filters designed for binaural HAD systems are referred to as binaural beamformers.

Binaural beamformers typically change the magnitude and phase relations of the microphone signals by forming a beam towards the target's direction while ideally suppressing all other directions. This may alter the spatial impression of the acoustic scene, as the filtered sources now reach both ears with possibly different relative phase and magnitude differences compared to before processing. This will appear unnatural to the HAD user. Therefore, there is an increasing interest in the preservation of the spatial information (also referred to as binaural cues) of the acoustic scene after processing. The present dissertation is mainly concerned with this particular problem and proposes several alternative binaural beamformers which try to exploit the available degrees of freedom to achieve optimal performance in both noise reduction and binaural-cue preservation.

Additional information ...


Signal Processing Seminar

audio signal processing, localization

Jie Zhang

Additional information ...


Signal Processing Seminar

Audio processing

Andreas Koutrouvelis

Additional information ...


Signal Processing Seminar

Array signal processing, Sensor networks, Optimization, Numerical Lineal Algebra

Mario Coutiño Minguez


Signal Processing Seminar

Distributed Convex Optimization: A Monotone Perspective

Thomas Sherson

Over the last few decades, methods of parallel and distributed computation have become essential tools in a wide range of applications such as machine learning, wireless sensor network processing and big data signal processing. Motivated by this point and the synergy between signal processing and convex optimization, in this work we demonstrate recent results in the area of distributed optimisation to facilitate such computation. In particular we highlight the primal dual method of multipliers (PDMM), a relatively recent algorithm proposed for distributed optimization. We demonstrate how PDMM can be derived from classic monotone operator theory which in turn provides insight into previously unknown convergence results for the algorithm. Using this insight we generalise PDMM to solve the class of separable problems with separable constraints and analyze how, in the case of strongly convex and smooth functions, the convergence rate of PDMM is influenced by the underlying network topology.

Additional information ...


MSc SS Thesis Presentation

Respiration monitoring based on information fusion from Impedance pneumography and Electrocardiography

Yuyang Wang

Wearable health has become a striking area in our daily life. Electrocardiogram (ECG) is one of the biomedical signals collected by the wearable or portable devices, which is widely used in heart rate monitoring and cardiac diagnosis. However, automatic ECG signal analysis is dicult in real application because the signals are easy to be contaminated by the noise and artifacts. Thus, the quality of ECG signals is essential for the accurate analysis.

The objective of this project is to design a reliable automated ECG signal quality indicator based on the supervised learning algorithm, which intends to estimate the quality of the signals and distinguish them. The methodology of this project is creating a classication model to indicate the quality of ECG signals based on the machine learning algorithm. The model is trained by the extracted features based on the Fourier transform, Wavelet transform, Autocorrelation function and Principal component analysis of ECG signals. Subsequently, the feature selection techniques are proposed to remove the irrelevant and redundant features and then the selected features are fed to classi- cation algorithms. The classier was then trained and tested on the expert-labeled data from the collected ECG signals. Particularly, we focus on the performance of classier and use the best training model to predict the quality of new ECG signals.


MSc BME Thesis Presentation

The effect of dopamine release on electrical neural activity in the prefrontal cortex

Jack Tchimino

How can certain oscillations be detected from the measured brain signals?

Additional information ...


Signal Processing Seminar

Tensor-based blind source separation in epileptic EEG and fMRI

Borbála Hunyadi

Additional information ...