Recent increase in widespread use of facial recognition technologies have accelerated the utilization of demographic information, as extracted from facial features, yet it is accompanied by ethical concerns. It is therefore crucial, for ethical reasons, to ensure that algorithms like face recognition algorithms employed in legal proceedings are equitable and thoroughly documented across diverse populations. Accurate classification of demographic traits are therefore essential for enabling a comprehensive understanding of other algorithms. This thesis explores how classical machine learning algorithms compare to deep-learning models in predicting sex, age and skin color, concluding that the more compute-heavy deep-learning models, where the best performing models achieved an MCC of 0.99, 0.48 and 0.85 for sex, age and skin color respectively, significantly outperform their classical machine learning counterparts which achieved an MCC of 0.57, 0.22 and 0.54 at best. Once establishing that the deep-learning models are superior, further methods such as semi-supervised learning, a multi-characteristic classifier, sex-specific age classifiers and using tightly cropped facial images instead of upper-body images were employed to try and improve the deep-learning results. Throughout all deep-learning experiments the state of the art vision transformer and convolutional neural network were compared. Whilst the different architectures performed remarkably alike, a slight edge was seen for the convolutional neural network. The results further show that using cropped facial images generally improve the model performance and that more specialized models achieve modest improvements as compared to their less specialized counterparts. Semi-supervised learning showed potential in slightly improving the models further. The predictive performances achieved in this thesis indicate that the deep-learning models can reliably predict demographic features close to, or surpassing, a human.

The ability of the visual system for object recognition is remarkable. A better understanding of its processing would lead to better computer vision systems and could improve our understanding of the underlying principles which produce intelligence. We propose a computational model of the visual areas V1 and V2, implementing a rich connectivity inspired by the neocortical circuit. We combined the three most important cortical plasticity mechanisms. 1) Hebbian synaptic plasticity to learn the synapse strengths of excitatory and inhibitory neurons, including trace learning to learn invariant representations. 2) Intrinsic plasticity to regulate the neurons responses and stabilize the learning in deeper layers. 3) Structural plasticity to modify the connections and to overcome the bias for the learnings from the initial definitions. Among others, we show that our model neurons learn comparable receptive fields to cortical ones. We verify the invariant object recognition performance of the model. We further show that the developed weight strengths and connection probabilities are related to the response correlations of the neurons. We link the connection probabilities of the inhibitory connections to the underlying plasticity mechanisms and explain why inhibitory connections appear unspecific. The proposed model is more detailed than previous approaches. It can reproduce neuroscientific findings and fulfills the purpose of the visual system, invariant object recognition. / Das visuelle System des Menschen hat die herausragende Fähigkeit zur invarianten Objekterkennung. Ein besseres Verständnis seiner Arbeitsweise kann zu besseren Computersystemen für das Bildverstehen führen und könnte darüber hinaus unser Verständnis von den zugrundeliegenden Prinzipien unserer Intelligenz verbessern. Diese Arbeit stellt ein Modell der visuellen Areale V1 und V2 vor, welches eine komplexe, von den Strukturen des Neokortex inspirierte, Verbindungsstruktur integriert. Es kombiniert die drei wichtigsten kortikalen Plastizitäten: 1) Hebbsche synaptische Plastizität, um die Stärke der exzitatorischen und inhibitorischen Synapsen zu lernen, welches auch „trace“-Lernen, zum Lernen invarianter Repräsentationen, umfasst. 2) Intrinsische Plastizität, um das Antwortverhalten der Neuronen zu regulieren und damit das Lernen in tieferen Schichten zu stabilisieren. 3) Strukturelle Plastizität, um die Verbindungen zu modifizieren und damit den Einfluss anfänglicher Festlegungen auf das Lernergebnis zu reduzieren. Neben weiteren Ergebnissen wird gezeigt, dass die Neuronen des Modells vergleichbare rezeptive Felder zu Neuronen des visuellen Kortex erlernen. Ebenso wird die Leistungsfähigkeit des Modells zur invariante Objekterkennung verifiziert. Des Weiteren wird der Zusammenhang von Gewichtsstärke und Verbindungswahrscheinlichkeit zur Korrelation der Aktivitäten der Neuronen aufgezeigt. Die gefundenen Verbindungswahrscheinlichkeiten der inhibitorischen Neuronen werden in Zusammenhang mit der Funktionsweise der inhibitorischen Plastizität gesetzt, womit erklärt wird warum inhibitorische Verbindungen unspezifisch erscheinen. Das vorgestellte Modell ist detaillierter als vorangegangene Arbeiten. Es ermöglicht neurowissenschaftliche Erkenntnisse nachzuvollziehen, wobei es ebenso die Hauptleistung des visuellen Systems erbringt, invariante Objekterkennung. Darüber hinaus ermöglichen sein Detailgrad und seine Selbstorganisationsprinzipien weitere neurowissenschaftliche Erkenntnisse und die Modellierung komplexerer Modelle der Verarbeitung im Gehirn.

Diese Dissertation untersucht die Syntax nichtkanonischer Subjekte und Subjektpositionen - insbesondere zwei Phänomene: Lokativinversion (LI) in Englisch, Französisch, Italienisch und Hebräisch; und Verbdrittverletzungen der Verbzweitregel (V2) in Kiezdeutsch - eine urbane Kontaktvarietät des Deutschen. In LI besetzt eine Lokativ-XP die präverbale Stelle, aber das kanonische DP-Subjekt in Nominativ taucht postverbal auf. Sprachübergreifend vergleiche ich a) die Verteilung unterschiedlicher Null- und overten Argumente in LI und b) die Verfügbarkeit von LI in Matrix- und Nebensätzen. Die zweite Fallstudie befasst sich mit kiezdeutschen V2-Verletzungen, denn sie folgen einer regelmäßigen Reihenfolge: [Rahmensetzer > Subjekt > finites Verb]; dies ist bemerkenswert aufgrund der Verletzung der ansonsten strengen Verbzweitregel und auch, weil es auf die Innovation einer Art Subjektposition hindeutet, die im Standarddeutschen fehlt. Anhand einer Korpusstudie komme ich zu der Erkenntnis, dass die scheinbare Subjektvoraussetzung auch für resumptive V3-Dislozierungsphänomene gilt. Dennoch wird gezeigt, dass die Subjektvoraussetzung mit dem nominativischen DP-Subjekten verbundenen EPP nicht ganz übereinstimmt. Ich entwickele eine Theorie von Subjektvoraussetzungen, die sowohl die Breite der untersuchten Variation in LI als auch die An/Abwesenheit von Subjektvorausetzungen im Vorfeld bei V3-Sätzen im Standarddeutschen und Kiezdeutschen erklären kann. Schließlich lassen sich diese Phänomene durch unterschiedliche Verteilungen und Vererbungsoperationen von D-, ϕ- und informationsstrukturellen δ-Merkmalen (cf. Miyagawa 2017) zwischen dem Phasenkopf C und T erklären. Die Anwesenheit nichtkanonischer Subjekte in LI und kanonischer Subjekte in einer nichtkanonischen Subjektposition im Kiezdeutschen werden durch Variation in der Verteilung eines für das Prädikatssubjekt spezifizierten δ-Merkmal abgeleitet. Dieses Merkmal ist aber unabhänging von den üblichen EPP-Voraussetzungen. / This dissertation addresses syntactic structures involving non-canonical subjects and non-canonical subject positions, investigating two phenomena: Locative Inversion (LI) in English, French, Italian, and Hebrew; and verb-third violations of the verb-second (V2) rule in Kiezdeutsch, an urban contact variety of German. In LI a spatio-deictic XP appears in the preverbal canonical subject position, while the canonical nominative subject DP surfaces postverbally. I compare the distribution of different covert and overt arguments participating in LI and the availability of LI in embedded and matrix contexts crosslinguistically. The second case study concentrates on Kiezdeutsch V2 violations, as they follow a regular order of [frame-setting adverb > Subject > finite verb]; this is remarkable because it both violates an otherwise strict V2 requirement and also indicates the innovation of a subject position lacking in Standard German. I carry out a corpus study and find that an apparent subject requirement extends to other verb-third resumptive-dislocation phenomena, yet we cannot understand this requirement in the sense of an EPP position associated with nominative DP subjects. From a theoretical perspective, this dissertation develops a theory of subject requirements capable of accounting for the breadth of investigated crosslinguistic variation in LI and the presence or absence of a high clausal subject requirement in Kiezdeutsch V2-violations and more standard varieties of German. Ultimately, I make use of finite differences across C and T in the distribution of D, ϕ, and discourse-related δ-features (cf. Miyagawa 2017) via different inheritance options from the phase head. The presence of non-canonical subjects in LI and the presence of canonical subjects in a seemingly non-canonical subject position in Kiezdeutsch are both derivable via variation in the placement of a δ-feature with a specification for Subject of Predication orthogonal to typical EPP requirements.

<p> This thesis presents work and simulations containing the use of Artificial Intelligence for real-time perception and real-time anomaly detection using the computer and sensors onboard an Unmanned Aerial Vehicle. One goal of this research is to develop a highly accurate, high-performance computer vision system that can then be used as a framework for object detection, obstacle avoidance, motion estimation, 3D reconstruction, and vision-based GPS denied path planning. The method developed and presented in this paper integrates software and hardware techniques to reach optimal performance for real-time operations. </p> <p>This thesis also presents a solution to real-time anomaly detection using neural networks to further the safety and reliability of operations for the UAV. Real-time telemetry data from different sensors are used to predict failures before they occur. Both these systems together form the framework behind the Intelligent UAV platform, which can be rapidly adopted for different varieties of use cases because of its modular nature and on-board suite of sensors. </p>

[ES] Los desórdenes del equilibrio y la marcha se encuentran entre los déficits motores más frecuentes entre aquellos individuos que han sufrido un ictus. En la clínica, estas habilidades son comúnmente evaluadas mediante herramientas clínicas que, pese a ser generalmente fáciles y rápidas de administrar, adolecen de poca precisión y estar sesgadas. Los sistemas instrumentados de laboratorio existentes para valorar la postura y la marcha resuelven potencialmente estas limitaciones a costa de requerir una preparación previa por parte de los evaluadores, un amplio espacio reservado en la clínica, un elevado tiempo de realización, y tener un coste muy elevado. El desarrollo tecnológico del sector del entretenimiento ha dado lugar en la última década a periféricos, como plataformas de presión y sensores de profundidad, que permiten la interacción mediante movimientos corporales manteniendo un bajo coste y una gran portabilidad y accesibilidad. Estudios iniciales han mostrado un rendimiento de estos dispositivos muy prometedor, y a veces comparable al de sistemas de laboratorio. Sin embargo, la falta de acceso a los sistemas desarrollados, la escasa investigación en personas con ictus y el desconocimiento de las propiedades psicométricas de las pruebas basadas en estos dispositivos en esta población comprometen la relevancia clínica que podrían tener estos sistemas. La hipótesis principal de este trabajo es que plataformas de fuerzas y sensores de profundidad de bajo coste, como la Nintendo Wii Balance Board y la Microsoft Kinect v2, respectivamente, pueden proporcionar información válida para cuantificar y evaluar la postura y la marcha de sujetos que han sufrido un ictus. Durante la presente tesis doctoral, por tanto, se ha llevado a cabo el desarrollo de herramientas de valoración de la postura y la marcha mediante los dispositivos nombrados, se ha posibilitado su acceso libre, se han determinado los valores normativos de las pruebas incluidas en las herramientas desarrolladas, y, finalmente, se ha investigado su sensibilidad, su validez convergente con herramientas clínicas estandarizadas y su fiabilidad inter e intraevaluador. Los resultados obtenidos de la participación de un total de 544 sujetos sanos y 173 sujetos con ictus en los cinco estudios que comprenden este trabajo evidencian que las herramientas desarrolladas permiten caracterizar satisfactoriamente la postura y la marcha de sujetos con ictus con respecto a la de sujetos sanos, poseen una validez convergente con instrumentos variables y coherente, y tienen una fiabilidad inter e intraevaluador excelente para casi todas las pruebas incluidas. Estos hallazgos sugieren que, pese a las limitaciones existentes, las herramientas desarrolladas podrían ser potencialmente usadas como una alternativa de bajo coste a los sistemas de laboratorio existentes para complementar la valoración de la postura y la marcha de sujetos con ictus. / [CA] Els desordres de l'equilibri i la marxa es troben entre els dèficits motors més freqüents entre aquells individus que han patit un ictus. A la clínica, aquestes habilitats són comunament avaluades mitjançant instruments clínics que, tot i ser generalment fàcils i ràpids d'administrar, poden tindre poca precisió i solen estar esbiaixades. Els sistemes instrumentats de laboratori existents per avaluar la postura i la marxa permeten resoldre aquestes limitacions, però requereixen una preparació prèvia per part dels avaluadors, un ampli espai reservat a la clínica, un elevat temps per realitzar cada prova i tenen un cost molt elevat. El desenvolupament tecnològic del sector de l'entreteniment ha donat lloc a plataformes de pressió i sensors de profunditat de baix cost, gran portabilitat i accessibilitat, que permeten la interacció amb entorns virtuals mitjançant moviments corporals. Estudis preliminars han mostrat un rendiment d'aquests dispositius molt prometedor i, de vegades, comparable al de sistemes de laboratori. No obstant això, la falta d'accés a les aplicacions desenvolupades, l'escassa investigació en persones amb ictus i el desconeixement de les propietats psicomètriques de les proves basades en aquests dispositius en aquesta població comprometen la rellevància clínica dels resultats obtinguts. La hipòtesi principal d'aquest treball és que plataformes de forces i sensors de profunditat de baix cost, com la Nintendo Wii Balance Board i la Microsoft Kinect v2, respectivament, poden proporcionar informació vàlida per quantificar i avaluar la postura i la marxa de subjectes amb ictus. Durant la present tesi doctoral, per tant, s'han desenvolupat eines de valoració de la postura i la marxa mitjançant els dispositius anomenats, s'ha possibilitat el seu accés lliure, s'han determinat els valors normatius de les proves incloses en les eines desenvolupades, i, finalment, s'ha investigat la seva sensibilitat, la seva validesa convergent amb instruments clínics estandarditzats i la seua fiabilitat inter i intraavaluador. Els resultats obtinguts de la participació d'un total de 544 subjectes sans i 173 subjectes amb ictus en els cinc estudis que comprenen aquest treball evidencien que les eines desenvolupades permeten caracteritzar satisfactòriament la postura i la marxa de subjectes amb ictus respecte a la de subjectes sans, tenen una validesa convergent amb instruments variable i coherent, i tenen una fiabilitat inter i intraavaluador excel·lent en gairebé totes les proves incloses. Aquestes troballes suggereixen que, tot i les limitacions existents, les aplicacions desenvolupades podrien ser potencialment usades com una alternativa de baix cost als sistemes de laboratori existents per complementar la valoració de la postura i la marxa de subjectes amb ictus. / [EN] Balance and gait disorders are common after stroke. In the clinical setting, these skills are usually assessed using clinical instruments that, despite being generally quick and easy to administer, may have limited accuracy and be biased. Instrumented laboratory-grade systems aimed at assessing posture and gait can potentially overcome these limitations. However, they require specific training to be operated and a long time to perform the assessments, and are usually bulky and expensive. In the last decade, the technological advances in the gaming industry have given rise to low-cost, portable and off-the-shelf devices, such as pressure platforms and depth sensors, which enable interaction with videogames through body movements. Previous research on the performance of these devices has shown promising results, and suggests that some measures could have comparable accuracy to those estimated by laboratory-grade systems. However, the lack of access to the software used in the experiments, the limited research in stroke patients, and the absence of knowledge about the psychometric properties of the assemment tests based on these devices, could limit the clinical relevance of the preliminary findings. The main hypothesis of this thesis is that low-cost force platforms and depth sensors, such as the Nintendo Wii Balance Board and the Microsoft Kinect v2, can provide sensitive, valid and reliable information to quantify and asses the postural control and gait of individuals with stroke, respectively. This work describes the development of two customized applications to assess posture and gait using the devices mentioned above, their publication on a dedicated website, the exploration of the normative values of tests included in the assessment, and, finally, the investigation of the sensitivity, convergent validity with standardized clinical instruments, and their inter- and intra-rater reliability. A total of 544 healthy subjects and 173 individuals with stroke have participated in the five studies that encompass this thesis. The results of these studies showed good sensitivity to motor impairment, variable and consistent convergent validity with clinical instruments, and excellent inter- and intra-rater reliability for almost all the tests examined. All these findings suggest that, despite their limitations, the developed applications interfaced with low-cost force platforms and depth sensors, could be potentially used as a low-cost alternative to instrumented laboratory-grade systems to complement the clinical assessment of the posture and gait of individuals with stroke. / Latorre Grau, J. (2022). Valoración del equilibrio y la marcha mediante sistemas de bajo coste en sujetos con ictus [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/181339

Multi functional health monitoring wearable devices are quite prominent these days. Usually these devices are battery-operated and consequently are limited by their battery life (from few hours to a few weeks depending on the application). Of late, it was realized that these devices, which are currently being operated at fixed voltage and frequency, are capable of operating at multiple voltages and frequencies. By switching these voltages and frequencies to lower values based upon power requirements, these devices can achieve tremendous benefits in the form of energy savings. Dynamic Voltage and Frequency Scaling (DVFS) techniques have proven to be handy in this situation for an efficient trade-off between energy and timely behavior. Within imec, wearable devices make use of the indigenously developed MUSEIC v2 (Multi Sensor Integrated circuit version 2.0). This system is optimized for efficient and accurate collection, processing, and transfer of data from multiple (health) sensors. MUSEIC v2 has limited means in controlling the voltage and frequency dynamically. In this thesis we explore how traditional DVFS techniques can be applied to the MUSEIC v2. Experiments were conducted to find out the optimum power modes to efficiently operate and also to scale up-down the supply voltage and frequency. Considering the overhead caused when switching voltage and frequency, transition analysis was also done. Real-time and non real-time benchmarks were implemented based on these techniques and their performance results were obtained and analyzed. In this process, several state of the art scheduling algorithms and scaling techniques were reviewed in identifying a suitable technique. Using our proposed scaling technique implementation, we have achieved 86.95% power reduction in average, in contrast to the conventional way of the MUSEIC v2 chip’s processor operating at a fixed voltage and frequency. Techniques that include light sleep and deep sleep mode were also studied and implemented, which tested the system’s capability in accommodating Dynamic Power Management (DPM) techniques that can achieve greater benefits. A novel approach for implementing the deep sleep mechanism was also proposed and found that it can obtain up to 71.54% power savings, when compared to a traditional way of executing deep sleep mode. / Nuförtiden så har multifunktionella bärbara hälsoenheter fått en betydande roll. Dessa enheter drivs vanligtvis av batterier och är därför begränsade av batteritiden (från ett par timmar till ett par veckor beroende på tillämpningen). På senaste tiden har det framkommit att dessa enheter som används vid en fast spänning och frekvens kan användas vid flera spänningar och frekvenser. Genom att byta till lägre spänning och frekvens på grund av effektbehov så kan enheterna få enorma fördelar när det kommer till energibesparing. Dynamisk skalning av spänning och frekvens-tekniker (såkallad Dynamic Voltage and Frequency Scaling, DVFS) har visat sig vara användbara i detta sammanhang för en effektiv avvägning mellan energi och beteende. Hos Imec så använder sig bärbara enheter av den internt utvecklade MUSEIC v2 (Multi Sensor Integrated circuit version 2.0). Systemet är optimerat för effektiv och korrekt insamling, bearbetning och överföring av data från flera (hälso) sensorer. MUSEIC v2 har begränsad möjlighet att styra spänningen och frekvensen dynamiskt. I detta examensarbete undersöker vi hur traditionella DVFS-tekniker kan appliceras på MUSEIC v2. Experiment utfördes för att ta reda på de optimala effektlägena och för att effektivt kunna styra och även skala upp matningsspänningen och frekvensen. Eftersom att ”overhead” skapades vid växling av spänning och frekvens gjordes också en övergångsanalys. Realtidsoch icke-realtidskalkyler genomfördes baserat på dessa tekniker och resultaten sammanställdes och analyserades. I denna process granskades flera toppmoderna schemaläggningsalgoritmer och skalningstekniker för att hitta en lämplig teknik. Genom att använda vår föreslagna skalningsteknikimplementering har vi uppnått 86,95% effektreduktion i jämförelse med det konventionella sättet att MUSEIC v2-chipets processor arbetar med en fast spänning och frekvens. Tekniker som inkluderar lätt sömn och djupt sömnläge studerades och implementerades, vilket testade systemets förmåga att tillgodose DPM-tekniker (Dynamic Power Management) som kan uppnå ännu större fördelar. En ny metod för att genomföra den djupa sömnmekanismen föreslogs också och enligt erhållna resultat så kan den ge upp till 71,54% lägre energiförbrukning jämfört med det traditionella sättet att implementera djupt sömnläge.