[diss] Sähkötekniikan korkeakoulu / ELEC

Permanent URI for this collectionhttps://aaltodoc.aalto.fi/handle/123456789/53

Browse

Recent Submissions

Now showing 1 - 20 of 1078
  • Dielectric materials for organic (light-emitting) transistors
    (2025) Gallegos Rosas, Katherine
    School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2025-07-18
    Organic semiconductors, characterized by van der Waals bonding, exhibit properties such as light weight, flexibility, and compatibility with low-temperature processing. These characteristics offer numerous advantages for large-scale and cost-effective manufacturing processes, rendering them highly relevant to the semiconductor industry and new-generation electronics. Currently, applications of organic devices are prevalent in displays, lighting, and solar cells, with organic light-emitting diodes leading the market. Following the development of organic light-emitting diodes, organic light-emitting transistors have emerged as promising devices that integrate light emission with the switching functionalities of transistors. Despite their potential to exceed the efficiency of diodes, fabricating high-performance light-emitting transistors requires precise material selection and optimization. A critical element in this context is the gate dielectric layer, which significantly impacts charge accumulation and charge transport at the interface with the organic semiconductor. This dissertation investigates distinct dielectric materials and their impact on the performance of organic (light-emitting) transistors. To this end, four types of dielectrics were evaluated: low-k polymer, high-k ferroelectric polymer, inorganic/organic bilayer, and biopolymer. By examining dielectric properties, this research aims to identify optimal material systems that can improve overall device performance, particularly regarding charge carrier transport and light-emission efficiency. The insights derived from this study are expected to contribute to developing advanced, high-performance electronic devices.
  • Fooling senses with electrical and magnetic stimulation of the retina and vestibular system
    (2025) Nissi, Janita
    School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2025-07-01
    Magnetic and electric stimulation of the sensory organs can cause artificial sensations even in the absence of a natural stimulus. For example, stimulation of the retina in the eye evokes visual sensations of white flickering lights called phosphenes, whereas stimulation of the vestibular system inside the ear evokes perceptions of movement and unstable balance. Various non-invasive stimulation methods have been used to study the function of the retina and the vestibular system. However, there is limited data on the internal electric fields induced by the stimulation that directly affect the sensory organs. This thesis consists of five peer-reviewed scientific publications. The publications characterise the properties of the retinal and vestibular electric fields that interfere with normal sensory function, produce phosphenes and alter balance. Publications I and II examine the electric field thresholds of phosphene perception in the retina evoked by external magnetic fields and transcranial currents, and the origin of the phenomenon. Publications III-V investigate the strength and frequency of the vestibular electric fields capable of causing lateral swaying while standing. The publications also assess the effect of the field direction and the possible sites of interaction within the vestibular system. The publications combine experimental measurements with state-of-the-art computational methods based on high-resolution anatomical models to provide realistic data on the strength and variability of the retinal and vestibular fields. The data can be used for the study of the underlying sensory mechanisms and the development of therapeutic sensory-stimulation applications aiming to help patients suffering from retinal or vestibular dysfunctions. Additionally, the findings of this work help to avoid the unwanted stimulation of these sensory organs, for example, during transcranial brain stimulation studies. The findings could also help to inform the revision of international health and safety guidelines that aim to protect humans from adverse reactions caused by low-frequency electromagnetic fields
  • Manipulation and assembly of objects using spatially nonlinear stochastic force fields
    (2025) Kopitca, Artur
    School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2025-07-01
    Spatially nonlinear stochastic force fields, omnipresent in nature, exhibit extraordinary capabilities for shaping, transporting, and assembling objects—yet they remain vastly underexplored and underutilized. For example, winds, with stochastic variations in speed and direction, gradually form sand dunes with complex shapes and transport various objects, from seeds to debris, over vast distances. Inspired by such natural phenomena, this thesis explores the potential of such forces for the controlled assembly and remote manipulation of objects through two exemplary fields: vibration fields and airflow fields. Despite differing in physical origin, these fields share common qualitative properties, including nonlinear spatial variations in force strength and direction, combined with random fluctuations. First, the thesis demonstrates that acoustic vibration fields can be programmed to assemble particles into desired two-dimensional shapes on a vibrating plate. Prior approaches to vibrationinduced particle assembly were constrained to shapes aligned with the intrinsic nodal patterns of the vibrating plate. This work extends these boundaries by employing data-driven models to predict stochastic particle motion and an optimization algorithm to iteratively minimize the gap between the desired and actual particle distributions on the plate. This enables the assembly of up to 100 particles into complex, recognizable shapes, such as Latin letters and geometric figures. Unlike conventional externally directed assembly methods that rely on static field- or template-based energy minimization, the proposed approach mimics natural shaping processes driven by long-term, timevarying, nonlinear external stimuli. Second, airflow fields are examined as tools for remote, non-contact manipulation of diverseobjects. Conventional non-contact techniques, such as magnetic, acoustic, or optical manipulation, are mostly limited by material specificity, strict shape requirements, or short operational ranges, typically confined to a few centimeters. In contrast, this thesis demonstrates that the airflow field induced by a single air jet can remotely and automatically manipulate objects of diverse materials and shapes over distances of up to 2.7 meters away, achieving a mean path-following error of 1.5 cm or less. This approach is shown to be effective on both solid and water surfaces, demonstrates robustness in the presence of obstacles and airflow disturbances, and is applicable to a range of practical scenarios. To achieve automatic object manipulation, two control strategies are introduced: a model-free controller, which operates purely on machine vision feedback, and a model-based controller, which leverages an analytical airflow field model and learned object dynamics, enabling multi-object control. This research addresses the gap between nature-inspired and engineering-driven manipulation and assembly methods. It underscores the transformative potential of spatially nonlinear stochastic force fields in manufacturing, robotics, and other applications, offering novel paradigms for fieldbased assembly and versatile, non-contact manipulation technologies.
  • Methods improving software design efficiency for flexible industrial automation
    (2025) Jhunjhunwala, Pranay
    School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2025-06-13
    The industrial automation world has shifted from mass production to mass customisation. Automation systems need to be flexible, modular, and efficient more than ever before and in order to support these paradigm shifts and changes, the need for software-defined automation has been drastically increasing. This thesis presents methods and tools that help improve the efficiency of software design for flexible industrial automation. The IEC~61499 standard for vendor agnostic distributed automation is used as the technology standard to develop the proposed methods. First, we discuss the need to optimise design efficiency when developing, debugging, testing, and maintaining industrial automation code. Different design patterns and software constructs are proposed that help increase reuse of software components, make systems more flexible and modular, and improve debugging and error handling capabilities. All contributing towards improving the software design efficiency. Next, we discuss the importance and effect of interoperability for flexible industrial automation systems. Vendors need to integrate state-of-the-art technologies in order to keep up with market changes. However, traditional automation systems, being rigid in nature, act as a barrier to higher levels of interoperability. We discuss the use of standardised interfaces that help achieve higher levels of interoperability and analyse its effect on overall system efficiency. Finally, in this thesis, we propose software driven reconfiguration of manufacturing systems aimed at reducing the downtime and error rates when handling systems that are dynamic in nature.
  • Control electronics for tuneable Fabry-Perot interferometers in hyperspectral imagers
    (2025) Holmlund, Christer
    School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2025-06-13
    Hyperspectral imaging has been rapidly evolving during the last decades. In hyperspectral imaging, a contiguous spectrum is acquired for each pixel in an image. Typically, a hyperspectral camera covers part of the wavelength range from the near ultraviolet to the mid-infrared in tens to hundreds of wavelength channels. This capability enables identifying and analysing materials and substances based on their spectral characteristics. Hyperspectral imagers or hyperspectral cameras have been used in satellites for various purposes, including environmental monitoring, atmospheric studies, and solar and astronomic research. A tuneable Fabry-Perot interferometer can be used as a scanning optical bandpass filter in a spectral imager. The transmitted wavelengths depend on the distance between the mirrors of the interferometer. In order to achieve the required spectral stability and repeatability, the mirror gap has to be controlled with sub-nanometre precision. The performance must remain unchanged over a wide operating temperature range. This thesis presents the control of the Fabry-Perot interferometers used in the hyperspectral imagers of the nanosatellites Aalto-1 and PICASSO. The electronics of the controller for the Fabry-Perot interferometers of the ultraviolet channel in the ALTIUS spectral imager are described in some detail. The ALTIUS instrument and satellite, currently under construction, are part of the European Space Agency’s Earth Watch programme.
  • Frequency-diverse phase holograms for millimeter- and submillimeterwave computational imaging
    (2025) Pälli, Samu-Ville
    School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2025-06-06
    Millimeter- and submillimeter-wave imaging systems have gained significant attention due to their unique capabilities in penetrating many non-metallic materials and providing detailed spatial resolution, making them ideal for security screening, industrial inspection, and medical diagnostics applications. However, at these wavelengths, traditional imaging systems often rely on costly and complex hardware, including large detector arrays and mechanical or electronic beamsteering mechanisms, which limit their scalability and affordability. This thesis addresses these challenges by focusing on designing, optimizing, and applying frequency-diverse phase holograms for computational imaging systems at millimeter- and submillimeter-wave frequencies. The research aims to develop simple and cost-efficient diffractive optical elements capable of encoding spatial information from a region of interest into frequencydiverse field patterns over a wide bandwidth, thus enabling advanced computational imaging independent of conventional mechanical or electronic beamsteering methods. The research is motivated by the need to create practical imaging systems that combine high resolution with reduced hardware complexity, leveraging advances in hologram design and neural network-based image reconstruction. The first part of this thesis introduces the design process, optimization, fabrication, and characterization of frequency-diverse phase holograms. These holograms utilize discretized phase modulation through quasirandom surface profiles to create spatially varying field patterns across a wide bandwidth. The hologram synthesis process incorporates physical optics simulations to optimize parameters such as the operation bandwidth, diffraction efficiency, and frequency diversity. Two fabricated holograms are designed for dual-band operation at WR-15 (50–75 GHz) and WR-3.4 (220–330 GHz) and a third hologram for a custom frequency band of 325–355 GHz. The latter employs spatial filtering methods to further enhance frequency diversity by reducing field intensity outside the RoI. The fabricated holograms are characterized using near-field measurements, demonstrating close to 50% diffraction efficiency and sufficient frequency diversity. The second part of the thesis focuses on two imaging systems utilizing the designed holograms: a vector network analyzer (VNA)-based system and a frequency-modulated continuous-wave (FMCW) radar system. The VNA-based system, operating at waveguide bands of 50–75 GHz and 220–330 GHz, is used to demonstrate the ability to reconstruct images of targets using a trained neural network with close to diffraction-limited accuracy. The effects of available bandwidth and spatial characteristics of the illuminating field on imaging accuracy are studied. The FMCW radar system, operating at 325–355 GHz, integrates an optimized hologram with the fast sweep rate of the radar and on-the-fly neural network-assisted image reconstruction. The system represents a development from laboratory-grade equipment to a more compact and standalone imaging system. The imaging capability is demonstrated by imaging a rotating target in real-time with a 60-Hz frame rate and good accuracy.
  • Drone localization in GNSS-denied environments: Addressing the impacts of hardware imperfections, RIS assistance, and jammer nullification
    (2025) Meles, Mehari Belay
    School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2025-05-30
    Advances in Global Navigation Satellite System (GNSS)-free drone localization and tracking are essential for reliable operations in environments where GNSS signals are unavailable or compromised. This thesis explores robust localization frameworks through measurement-based methods for two-dimensional (2D) and three-dimensional (3D) localization, and simulation-based studies that assess hardware imperfections and jamming mitigation, with a focus on the potential of Reconfigurable Intelligent Surface (RIS) technology to enhance performance and robustness. The first part of the research explores a measurement-based 2D GNSS-free drone-localization framework, employing unique methodologies to establish a baseline for GNSS-free positioning. Building on this foundation, the thesis progresses to measurement-based 3D GNSS-free localization frameworks, utilizing Multiple-Input Multiple-Output (MIMO) testbeds with refined methods to enhance accuracy. This work lays a solid foundation by leveraging real-world measurements to demonstrate the feasibility and effectiveness of GNSS-free drone localization, while addressing practical challenges and proposing solutions in both 2D and 3D scenarios. Moving on to simulationbased approaches, the study evaluates the impact of hardware impairments, such as Carrier Frequency Offset (CFO), Phase Noise (PN) and tilting base station (BS) antennas which are crucial for understanding how hardware imperfections degrade the accuracy and reliability of GNSS-free drone localization under real-world conditions. The thesis further explores RIS-assisted GNSS-free drone localization as a solution in scenarios where cellular BSs are unavailable, demonstrating the potential of the solution to enhance localization performance and mitigate jamming. Finally, the research integrates RIS-assisted 3D GNSS-free drone localization with jamming nullification strategies, providing a comprehensive approach for secure and precise GNSS-free drone positioning. Together, these contributions establish a solid foundation for advancing secure and accurate drone localization methodologies, combining practical measurement insights with state-of-the-art simulation capabilities.
  • Machine learning and state-space methods for healthcare, speech, and maritime awareness
    (2025) Gorad, Ajinkya
    School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2025-05-30
    The aim of the thesis is to investigate state-of-the-art methods for isolated speech recognition using spiking neural networks, non-contact respiratory monitoring using thermal imaging, and multi-sensory approaches for ship and sea ice detection. The applications include methods for sound based ship foghorn bearing estimation, bearing estimation in visible and thermal infrared imaging, sea ice-track semantic detection for navigation, and generic multi object tracking. Although these applications span diverse domains, they are unified by the core challenge of extracting robust, structured information from noisy, time-varying sensory data. We explore architecture of liquid state machines to classify speech patterns. Using spiking neurons and models of cochlear processing, we introduce a novel performance measure, named memory metric 𝜏𝑀 to evaluate system's ability to classify speech. We non-invasively analyze breathing patterns by extracting airflow signals from nasal temperature changes captured with a thermal camera. Auditory and thermal signals are both temporally processed. In maritime awareness, we use sound sensors to estimate ship bearing, and visible and thermal infrared vision sensors are combined with deep learning and Kalman filters for multi-modal ship tracking. For navigating icy regions, we apply deep learning to identify sea ice in images, using fused visible and thermal imagery. Together, these domains highlight how data fusion, tracking, and classification can be applied across vision, acoustics, and thermal sensing. On the algorithm front, this thesis introduces a method for parameter estimation in extended Kalman filters and a deep Rao-Blackwellized particle filter. In the extended Kalman filter, automatic differentiation provides a robust way to obtain gradients with respect to the parameters for parameter estimation. We utilize this method to obtain model and noise parameters from data. Additionally, we introduce a deep Rao-Blackwellized Monte Carlo data association particle filter for multi-object tracking for image sequences. ReID features capturing appearance and shape are tracked alongside position and velocity. These algorithmic contributions demonstrate how principled probabilistic modeling and differentiable estimation techniques can generalize across distinct application settings. This thesis contributes by extensive data collection from ships during winter such as MS Megastar and icebreaker Sampo, lab studies with volunteers for non-contact breathing measurements, and laboratory measurements using microphone array. Results are supplemented by online datasets for speech and appearance-based object tracking. The results demonstrate the feasibility of using thermal cameras for breath monitoring as well as applicability to vision-based methods for ship tracking and multi-modal awareness in maritime settings. By bridging domains through shared computational strategies, the thesis highlights the potential for unified sensory modeling in realworld, data-driven environments.
  • Remote sensing of terrestrial snow water equivalent using satellite-based radiometer sensors
    (2025) Venäläinen, Pinja
    School of Electrical Engineering | Doctoral thesis (article-based)
    Seasonal snow cover is an important component of the Earth's hydrological and energy cycles, affecting water resources and climate feedback mechanisms. Snow water equivalent (SWE), representing the water content of a snowpack, is a key characteristic of snow cover. SWE estimates can be retrieved from passive microwave radiometer data. Global satellite-based passive microwave radiometer measurements are available from 1978 onwards allowing construction of long SWE time series. Radiometer-based SWE retrievals can be improved with the assimilation of synoptic snow depth observations. This thesis aims to advance assimilation-based SWE retrieval method with parametrization of snow density and bias correction. Publication I presents a method for creating climatological spatially and temporally dynamic snow density fields. The effect of post-processing SWE retrieval with these fields is also studied in the publication. Post-processing improves the overestimation of small SWE values and small improvements in the underestimation of large SWE values are also present. Publication II investigates implementing dynamic snow densities into the SWE retrieval. Similarly to post-processing with dynamic snow densities, implementing them into the retrieval improves the accuracy of (small) SWE estimates. Additionally, the reduction in hemispheric peak snow mass seen when post-processing with dynamic snow densities is smaller when snow densities are implemented into the SWE retrieval. Implementation of dynamic snow densities into SWE retrieval also delays peak snow mass timing and thus improves the seasonal evolution of SWE. Publication III updates previously studied monthly bias correction method for monthly SWE estimates with new reference data. Monthly bias correction is also expanded to a daily time scale in this publication. Updated monthly bias correction improves monthly estimates and daily bias correction slightly improves the accuracy of large SWE estimates. More importantly adds a significant amount of snow to the hemispheric snow mass estimation. Together these three studies improve SWE estimations and our ability to monitor seasonal snow cover.
  • Spoken Language Understanding: Deep Neural Network Approaches for Low-Resource Languages
    (2025) Porjazovski, Dejan
    School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2025-05-09
    Spoken language understanding (SLU) enables human-computer interaction through voice commands, powering applications like virtual assistants integrated into smartphone devices and vehicles. Advances in computational power and data availability have made automatic speech understanding feasible. However, SLU systems require substantial training data, which is available for high-resource languages like English but often scarce for low-resource languages like Finnish, leading to subpar performance. In this thesis, I will explore methods for enhancing SLU in low-resource settings, focusing on deep learning-based solutions. Traditional SLU systems employ a cascading architecture, where an automatic speech recognition (ASR) system generates transcripts processed by a natural language understanding (NLU) module. While effective, these systems face significant challenges in low-resource scenarios, where an ASR system may be unavailable or produce low-quality transcripts, propagating the errors to the NLU module. Moreover, separate training of ASR and NLU models results in the ASR model not being optimised for the NLU task and vice-versa. To address these issues, this thesis investigates neural SLU approaches that jointly optimise or eliminate the need for ASR. These approaches are applied to a diverse range of SLU tasks, including named entity recognition, topic identification, spoken emotion recognition, and request identification, across multiple languages such as Finnish, English, Swedish, and French, among others. A critical component of SLU is extracting meaningful representations from audio, known as embeddings. Given the abundance of speech embedding methods, selecting the optimal embedding for a specific task and language is non-trivial. This work proposes four principles for evaluating speech embeddings and presents extensive experiments adhering to these principles. Lastly, the thesis examines the generalisation of SLU models - a capability that allows the models to make correct predictions when presented with data distribution not seen during training. Robust generalisation to out-of-distribution data is crucial for real-world deployment, yet machine learning models, including SLU systems, often struggle. To facilitate research in this area, the thesis introduces SLU data splits designed to test models’ generalisation and provides insights into the limitations of current approaches. The contributions outlined in the thesis advance the field of SLU, especially for low-resource languages. These findings provide a solid foundation for robust SLU systems designed to work under diverse language and application constraints.
  • Modeling and Measurements of Human Body Effects on Millimeter-Wave and Sub-Terahertz Handset Antenna Radiation: From Permittivity Estimation to Spherical Coverage
    (2025) Xue, Bing
    School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2025-04-17
    As the demand for high data-rate telecommunications continues to grow, the limitations of the sub-6 GHz band become increasingly apparent, particularly due to its constrained bandwidth. In response, fifth-generation (5G) millimeter-wave (mmW) and sub-terahertz (sub-THz) frequencies have emerged as promising solutions, offering unprecedented data rates thanks to their wide bandwidth. However, electromagnetic waves at these higher frequencies experience more significant free-space path loss and are more easily obstructed by blockages, presenting substantial challenges in maintaining reliable, low-latency wireless channels. To understand and predict how wireless signals behave in complex environments, modeling wireless channels is essential. A key challenge in modeling these channels at high frequencies is accurately representing user-induced blockages, such as the effects of a user’s body and hands on handset antennas. Hand blockages involve complex near-field interactions with handset antennas, often requiring full-wave simulations for accurate analysis. Although these simulations are effective for modeling body blockages, they are highly time-consuming. The plane wave assumption, typically valid for free-space electromagnetic waves, does not apply to electromagnetic waves traveling through body blockages, which fall within the Fresnel region of handset antennas. Consequently, developing a reliable, computationally efficient model of human-antenna interactions for wireless channel modeling is critical. In the 5G mmW bands, this thesis presents a detailed study of hand effects on handset antennas. Measured radiation patterns using real hands and hand phantom models validate full-wave simulation models of hand-antenna interactions. Additionally, an analytical model based on knifeedge diffraction and geometric optics is introduced to estimate user body effects, offering a faster alternative to full-wave simulations with comparable spherical coverage predictions. In the sub-THz bands, novel permittivity characterization methods for thin and thick materials address phase calibration challenges in free-space measurements. Human skin permittivity is characterized using open-ended waveguides for small areas and free-space reflection coefficients for larger areas. These human skin permittivity data support the development of simulation models for analyzing hand effects on sub-THz handset antennas. Radiation pattern measurements using the proposed reference antennas and real hands validate the hand-antenna simulation models. Looking forward, these validated models provide a foundation for future research on hand effects in handset antenna design and wireless channel modeling. The proposed measurement methodologies will also support further experimental studies on hand-antenna interactions and material permittivity characterization, contributing to more accurate modeling of wireless channels at high frequencies.
  • Studies on antenna arrays: Advanced manufacturing methods and integration of microwave components
    (2025) Kuosmanen, Matti
    School of Electrical Engineering | G5 Artikkeliväitöskirja | Defence date: 2025-04-11
    Radars, electronic warfare devices, and telecommunication systems use antenna arrays to transmit and receive radio frequency (RF) signals. Compared to single antennas, antenna arrays can electrically adjust their radiation patterns. New frequency bands have constantly been allocated to these systems, and the bandwidth of the RF signals has been increasing. For example, typical military radars use frequencies from 150 MHz to 18 GHz, depending on the specific radar. Furthermore, the current fifth-generation (5G) mobile networks can use frequency bands from 410 MHz to 71 GHz. The constant evolution of these systems has led to stricter demands for antenna arrays. Better electrical performance, smaller size, and lower price are expected by the industry, and the antenna arrays should also be compatible with various platforms. This thesis proposes solutions for these challenges by focusing on two aspects: new manufacturing methods and novel integration techniques of microwave components into antenna arrays. The first part of this thesis presents new manufacturing techniques for making lightweight antenna arrays without compromising electrical performance. The antenna arrays designed have been manufactured using an unconventional method, in which the antenna elements are metalized cavities in foam or plastic material instead of solid metal. This so-called inverted manufacturing technique provides a material-efficient manner to fabricate antenna arrays, resulting in up to 73% mass reduction compared with conventional, all-metal arrays. The designed antenna arrays operate at 2–6 GHz and 6–18 GHz frequency ranges, and they can steer the beam ±50° in both planes. The second part of this thesis concentrates on the integration of microwave components into antenna arrays. The thesis presents two antenna arrays, where wideband band-pass filters and lowpass filters are incorporated into Vivaldi antenna elements. The filters are based on corrugated slotlines, which are strongly dispersive. The designed antenna arrays operate at 1.3–3.1 GHz and 6–18.5 GHz frequency ranges, and their out-of-band suppression is 20–40 dB. The performance metrics of the filters, namely their attenuation and cut-off frequency, are independent of the beamsteering angle, which is a desirable property for phased arrays. In addition to filtering antenna arrays, the thesis investigates the integration of wideband push–pull amplifiers into an antenna array. The active antenna array is designed employing an antenna–amplifier codesign methodology, where the antenna array is designed to provide a suitable load impedance for the amplifier without a separate matching network. Because both the antenna and amplifier are differential, a balun is not needed in the antenna–amplifier interface. The advantages of the proposed design are its small size, simple structure, and large 2–5 GHz bandwidth. Every active antenna element can produce up to 12 W power, and thus the array is suitable for high-power applications such as multi-static radar transmitters.
  • Crowdsourced 3D semantic mapping and change detection in urban driving environments
    (2025) Zhanabatyrova, Aziza
    School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2025-03-24
    The advancement of autonomous driving hinges on the availability of accurate and up-to-date semantic maps, which provide a detailed representation of the environment for safe and efficient route planning, obstacle avoidance, and decision-making in urban environment. In order to manage the dynamic nature of these environments, which are frequently modified by changes such as updated traffic signs, traditional methods of generating and updating semantic maps, involve specialized vehicles equipped with high-precision sensors, which are both time-consuming and costly. To mitigate these challenges, this dissertation investigates the potential of leveraging crowdsourced image data from consumer-grade cameras, such as smartphones and dashboard cameras, as well as traffic flow data, which can be obtained from existing mapping services like HERE Maps. In this dissertation, we first introduce a novel pipeline designed to automatically detect changes (e.g., types and locations of traffic signs) in complex, large-scale urban driving environments, utilizing point clouds generated from visual data using Structure from Motion (SfM). The pipeline comprises several components, each addressing specific challenges such as online change detection, and accurate change localization in 3D space. Second, we provide key guidelines for constructing large-scale 3D maps using visual crowdsourced data to ensure both accuracy and reliability. We examine the challenges posed by the inherent monocular nature and data inconsistency of visual crowdsourcing data, and conduct a comprehensive comparison of various SfM techniques on such data in complex, large-scale urban environments. Finally, we propose a deep-learning method for coarse-grained change detection using traffic flow data to reduce the costly and extensive search space required to detect changes in large-scale environments. In practice, this coarse-grained change detection can be used to initially identify areas of change, which can then be refined using the pipeline from the first step for precise localization and semantic map updates. The results of this work provide the basis for the future deployment of cost-effective automatic change detection for accurate and up-to-date semantic maps for autonomous driving.
  • Perceptual Effects of Sound Field Reproduction Methods within the context of Head-worn Microphone Arrays
    (2025) Fernandez, Janani Cheryl
    School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2025-04-01
    People who are prescribed hearing assistive devices, or use other head-worn hear-through devices, perceive the audible world differently compared to how they would normally experience the world directly with their ears. With the advent of augmented, mixed, and virtual reality devices, which are rapidly appearing on the consumer market, such head-worn hear-through devices are also being used by an increasing population of people who don't have hearing loss. Understanding the spatial aspects of the audible world around us is an important objective of human hearing, and the effect that such devices have on spatial perception is an active topic of study. However, the setups used in such studies can be prohibitive due to their cost and space requirements; as they typically require large loudspeaker installations. One approach to mitigate these drawbacks is to instead deploy spatial audio technologies, such as virtual audio rendering techniques, for evaluating the sound source localisation abilities of a listener; since such techniques can typically operate using far fewer loudspeakers. Sound field reproduction technologies can also be utilised to deliver entire (360 degree) sound scenes over similar loudspeaker setups, which device wearers can then experience under controlled laboratorysettings. Therefore, the first contributions of this thesis explored the feasibility of using virtual audio technlogies to test sound source localisation abilities of people wearing cochlear implants and hearing aids. Additionally, the use of certain sound field reproduction methods was explored, which aim to faithfully recreate sound scenes within clinical environments. Such reproduction methods may be used when configuring such hear-through devices, in order to ensure that they perform optimally for a variety of realistic sound scene conditions. They may also be used for testing the users' hearing abilities and the performance of the devices themselves. Spatial audio technologies may also be integrated within the devices themselves, in order to improve the delivery of spatial information to the listener; since spatial information can often be distorted by traditional speech intelligibility enhancement algorithms. Therefore, the final contributions of this thesis involved exploring signal processing techniques, which aim to preserve the spatial characteristics of the reproduced sound scene delivered by such devices. The first technology is a modification of a signal-independent reproduction method, which allows for a direction-dependent loudness control to be integrated into the rendering. The second technology is a framework that supports parametric spatial audio reproduction of the head-worn microphone array signals. The validation tests carried out in this study showed that by adopting a sound-field model, and employing signal-dependent rendering strategies, the perceived spatial accuracy of the reproduced sound scene can be improved.
  • Development and implementation of measurement principles and devices from single-photon applications to telecom
    (2025) Vaigu, Aigar
    School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2025-03-21
    This thesis addresses the challenge of traceably measuring properties of low-photon flux sources and detectors in optical radiometry, a critical task for low-photon and single-photon applications. The difficulty lies in the reliable detection and characterisation of low-photon fluxes at power levels, where conventional methods face detection limits. To solve this, the thesis presents a demonstration of a predictable single-photon source based on an optically excited silicon-vacancy centre in a nanodiamond. Operating at an excitation rate of 70 MHz, this source emits photon flux at 685 nm, sufficient to be measured by a low-optical-flux detector. By directly measuring the photon flux, the efficiency of the single-photon source is determined, enabling it to serve as an absolute reference for calibration. The original contribution of this work lies in the predictable control of the photon flux through the adjustment of the excitation laser’s repetition rate, achieving predictable photon flux levels below the detection limit of conventional low-optical-flux detectors. This method does not require prior knowledge of the source efficiency, as it can be calibrated using the low-optical-flux detector, offering a novel approach for the characterisation of detectors at the few-photon level. The thesis also introduces innovative designs for transmission trap detectors. Unlike conventional trap detectors, where photodiodes are wired in parallel, the photodiodes in these designs are independently monitored, providing additional data on incident light and detector linearity. Firstly, the collinear, polarisation-insensitive transmission trap detector, which uses eight silicon photodiodes, has transmittance lower than 10-4 and uniform spectral responsivity within 0.1% across a 4 mm × 4 mm area, representing a significant contribution in the field of trap detector research. Furthermore, a polarisation-insensitive, two-element transmission trap detector consisting of InGaAs photodiodes in a non-collinear configuration has a low spatial non-uniformity of less than 0.5% across its active area of over 9 mm in diameter. In contrast to conventional designs, the transmittance of this detector is adjusted not by the number of photodiodes but by the incidence angle of reflection from the photodiodes. Finally, the thesis introduces a hybrid two-element transmission trap detector combining Si and InGaAs photodiodes, which offers broad wavelength coverage from 400 nm to 1600 nm for telecommunications applications.
  • Channel Charting-based Radio Resource Management
    (2025) Kazemi, Parham
    School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2025-03-17
    5th Generation (5G) cellular networks are designed to deliver unparalleled performance in mobile environments with three promises: i) increased capacity, ii) ultra-reliable and low-latency connections, and iii) a massive number of connected devices. Achieving these ambitious goals necessitates the integration of novel technologies into networks. Millimeter-wave (mmWave) communications stand out as a key enabler for achieving ultrahigh data rates and low latency, leveraging the substantial bandwidth available at these high frequencies. Beamforming techniques have been extensively employed in the mm-wave bands to alleviate the path loss of mm- wave radio links. However, several challenges must be overcome, primarily associated with the high overhead of finding suitable beams. This thesis addresses key challenges in beam management for 5G and mmWave communication systems through the application of Channel Charting (CC) and Machine Learning (ML) techniques. CC is a self-supervised method that maps the collected high dimensional Channel State Information (CSI) at a Base Station (BS) into a low dimensional space which represents pseudo positions of User Equipment (UEs) in the radio environment. The low dimensional space preserves the local geometry of the UEs meaning that nearby UEs in real space are close to each other on the CC. A CC-based framework is designed where in an offline training phase, CCs are constructed and annotated with Signal-to-Noise Ratio (SNR)s of neighboring cells/beams. ML algorithms are used to predict the SNR of a user at neighboring cells/beams from its transmission in a massive Multiple Input Multiple Output (mMIMO) cellular system. By predicting the signal quality of neighboring stations without UE assistance, the protocol overhead for handover decisions can be reduced. Both standalone and non-standalone 5G system deployments are considered and the best beam prediction is investigated. Beam tracking based on CC is investigated and results show that at a very low beam-search overhead one can leverage a CCto-SNR mapping in order to track strong beams between the UEs and the BS. As the fundamental building block of the framework proposed in this thesis, CC necessitates enhancements in its construction to enable versatile applications across different scenarios. To address this, a CSI feature has been devised aimed at mitigating the influence of small-scale fading. This improvement empowers the framework to yield robust predictions even with low spatial sampling density. Additionally, a low complexity Out-of-Sample (OOS) algorithm has been developed, which boasts reduced computational requirements compared to conventional OoS algorithms, making it a more efficient choice for practical implementations.
  • Reflectance metrology of thin films
    (2025) Danilenko, Aleksandr
    School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2025-03-07
    This thesis investigates the application of reflectometry techniques for characterizing thin-film structures, with a focus on developing methods to analyse complex multilayer configurations. Accurate reflectance data is critical in reflectometry, therefore, the Richardson-Lucy bandwidth correction method was applied and tested on reflectance measurements of a 2-μm-thick SiO₂ layer on a silicon substrate. This evaluation assessed improvements, effects and artifacts the method might introduce on processed measured data. In parallel, a basic model of the PillarHall chip, featuring an air gap layer, was developed and analysed. This model employs a dedicated MATLAB code based on the transfer-matrix method to simulate the reflectance spectrum and fit it to the measured data, enabling the extraction of layers’ thicknesses of the studied sample. Building on the basic model, an advanced PillarHall model was created, designed not only to estimate layer thickness but also to determine the size of specific structural features. The findings underscore the importance of accurate, bandwidth-corrected reflectance spectra, as the precision of fitted model parameters correlates directly with the spectrum quality.
  • Low-temperature solid-liquid interdiffusion bonding for heterogeneous integration
    (2025) Golim, Obert Pradipta
    School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2025-02-28
    There has been a notable transition in the electronics industry towards the heterogeneous integration strategy to achieve a compact yet multifunctional devices. However, this approach present challenges when incorporating materials with differing characteristics, particularly regarding temperature-induced constraints for the assembly process. Low-temperature bonding techniques are one of the key technologies to address this problem and fully realize the potential of heterogeneous integration. Among the numerous bonding methods, this dissertation explores the low-temperature bonding processes based on the solid–liquid interdiffusion (SLID) technique utilizing Cu–Sn–In metallurgy. The eutectic behaviour of Sn–In allows a significant reduction of the bonding temperatures from a typical bonding temperature above 250 °C down to 150 °C. Microstructural characterization of the bonds formed at temperatures between 150 °C and 200 °C reveals a homogeneous interconnect featuring an intermetallic phase with a remelting temperature above 450 °C, which makes it thermally stable. Electrical measurements confirm the interconnects’ small resistance that is comparable to those obtained from well-known Cu–Sn bonding technique, underscoring the capability of this low-temperature process for vertical integration. Hermeticity assessment, conducted using seal-ring shaped bond, highlight the technique’s potential for microelectromechanical system (MEMS) or micro-optoelectromechanical system (MOEMS) packaging. Moreover, the bond exhibited tensile strength of a similar magnitude with the popular Cu–Cu thermocompression bonding method. Some challenges in the implementation of the low-temperature bonding process have been identified, including limitations for process integration, squeeze-out, and defect formations. These issues can be mitigated through optimization of the design and processing parameters. Despite these challenges, the results from this work highlight the prospect to fully demonstrate the capabilities of heterogeneous integration by addressing temperature-induced limitations.
  • Velvet noise in audio processing
    (2025) Fagerström, Jon
    School of Electrical Engineering | G5 Artikkeliväitöskirja
    Noise plays a central role in various audio-processing applications, including artificial reverberation, audio decorrelation, acoustical measurements, sound synthesis, and speech processing. This dissertation focuses on applications of sparse noise, known as velvet noise. With its minimal density and smooth temporal envelope, velvet noise has been widely used in artificial reverberation algorithms, both as a core component as well as a decorrelating or diffusing element. The work within builds on this foundation by exploring novel variants and applications of velvet noise. The thesis introduces dark velvet noise as a low-passed variant, generalizes this to extended dark velvet noise for accurate modeling of non-exponential late-reverberation, and culminates in developing the binaural dark-velvet-noise reverberator. Additionally, short velvet noise filters are explored for decorrelation and variation filtering tasks, demonstrating their effectiveness in lowering inter channel correlation within feedback delay networks through velvet feedback matrices and generating realistic variations of sampled percussive sounds. The contributions of this thesis offer significant advances in applying velvet noise to audio processing, with particular emphasis on artificial reverberation and decorrelation.
  • Semiconductor nanowires on flexible plastic substrates
    (2025) Khayrudinov, Vladislav
    School of Electrical Engineering | Doctoral thesis (article-based)
    Nanowires (NWs) offer exceptional potential for use in solar cells, lasers, LEDs, and photodetectors. In parallel, there's growing interest in flexible electronics due to their cost-effectiveness, lightweight design, mechanical resilience, and chemical stability. However, a major obstacle to integrating NWs with flexible electronics is the high temperatures typically required for metalorganic vapour phase epitaxy (MOVPE) growth, which makes NWs incompatible with most plastic substrates. This dissertation presents innovative techniques for the direct growth of III-V nanowires on flexible plastic substrates, along with methods to create functional nanowire-based flexible electronic devices. Firstly, this work establishes an isolated growth regime for self-catalysed InAs NWs, which enables their growth at record low temperatures. Extensive characterization reveals the ability to control crystal structure and NW density, which is particularly essential for achieving compatibility with lowtemperature processes necessary for flexible electronics. Next, the direct growth of III-V NWs on flexible plastic substrates is achieved. High-density, well-crystallized InAs and InP nanowires are grown on polyimide without pre-treatment, demonstrating strong mid-infrared emission for InAs and near-infrared emission for InP. Significantly, the electrical properties of these NWs allow for the fabrication of flexible nanowire-based p-n junction devices on plastic in just two fabrication steps. Additionally, this research marks the first successful growth of InSb nanowires on flexible plastic substrates. These NWs display high material quality, room temperature photoluminescence, and remarkable flexibility, showing promise for future flexible optoelectronics. Finally, a method for fabricating GaAs NW LEDs is presented. GaAs NWs are grown directly on flexible plastic substrates within the MOVPE reactor, showing zinc blende crystal structure, room temperature photoluminescence emission, and desirable optical properties. These findings open up possibilities for roll-to-roll compatible LED production using GaAs NWs, a key step in integrating these materials into a wide range of flexible optoelectronic devices. These advancements push the boundaries of flexible optoelectronics by enabling the direct growth of III-V nanowires on plastic substrates, while also lowering fabrication complexity and temperature requirements. The methods presented here pave the way for more versatile, durable, and efficient devices across a range of applications