Swarm robotics is the study of developing and controlling large groups of robots. Collectives of robots possess advantages over single robots such as being robust to mission failures due to single-robot errors. Experimental research in swarm robotics is currently limited by swarm robotic technology. Current swarm robotic systems are either small groups of sophisticated robots or large groups of simple robots due to manufacturing overhead, functionality-cost dependencies, and their need to avoid collisions, amongst others. It is therefore useful to develop a swarm robotic system that is easy to manufacture, that utilises its sensors beyond standard usage, and that allows for physical interactions. In this work, I introduce a new type of low-friction locomotion and show its first implementation in the HoverBot system. The HoverBot system consists of an air-levitation and magnet table, and a HoverBot agent. HoverBots are levitating circuit boards which are equipped with an array of planar coils and a Hall-effect sensor. HoverBot uses its coils to pull itself towards magnetic anchors that are embedded into a levitation table. These robots consist of a Printed Circuit Board (PCB), surface mount components, and a battery. HoverBots are easily manufacturable, robots can be ordered populated; the assembly consists of plugging in a battery to a robot. I demonstrate how HoverBot's low-cost hardware can be used beyond its standard functionality. HoverBot's magnetic field readouts from its Hall-effect sensor can be associated with successful movement, robot rotation and collision measurands. I build a time series classifier based on these magnetic field readouts, I modify and apply signal processing techniques to enable the online classification of the time-variant magnetic field measurements on HoverBot's low-cost microcontroller. This method allows HoverBot to detect rotations, successful movements, and collisions by utilising readouts from its single Hall-effect sensor. I discuss how this classification method could be applied to other sensors and demonstrate how HoverBots can utilise their classifier to create an occupancy grid map. HoverBots use their multi-functional sensing capabilities to determine whether they moved successfully or collided with a static object to map their environment. HoverBots execute an "explore-and-return-to-nest" strategy to deal with their sensor and locomotion noise. Each robot is assigned to a nest (landmark); robots leave their nests, move n steps, return and share their observations. Over time, a group of four HoverBots collectively builds a probabilistic belief over its environment. In summary, I build manufacturable swarm robots that detect collisions through a time series classifier and map their environment by colliding with their surroundings. My work on swarm robotic technology pushes swarm robotics research towards studies on collision-dependent behaviours, a research niche that has been barely studied. Collision events occur more often in dense areas and/or large groups, circumstances that swarm robots experience. Large groups of robots with collision-dependent behaviours could become a research tool to help invent and test novel distributed algorithms, to understand the dependencies between local to global (emergent) behaviours and more generally the science of complex systems. Such studies could become tremendously useful for the execution of large-scale swarm applications such as the search and rescue of survivors after a natural disaster.

Improving speech recognition performance in the presence of noise and interference continues to be a challenging problem. Automatic Speech Recognition (ASR) systems work well when the test and training conditions match. In real world environments there is often a mismatch between testing and training conditions. Various factors like additive noise, acoustic echo, and speaker accent, affect the speech recognition performance. Since ASR is a statistical pattern recognition problem, if the test patterns are unlike anything used to train the models, errors are bound to occur, due to feature vector mismatch. Various approaches to robustness have been proposed in the ASR literature contributing to mainly two topics: (i) reducing the variability in the feature vectors or (ii) modify the statistical model parameters to suit the noisy condition. While some of those techniques are quite effective, we would like to examine robustness from a different perspective. Considering the analogy of human communication over telephones, it is quite common to ask the person speaking to us, to repeat certain portions of their speech, because we don't understand it. This happens more often in the presence of background noise where the intelligibility of speech is affected significantly. Although exact nature of how humans decode multiple repetitions of speech is not known, it is quite possible that we use the combined knowledge of the multiple utterances and decode the unclear part of speech. Majority of ASR algorithms do not address this issue, except in very specific issues such as pronunciation modeling. We recognize that under very high noise conditions or bursty error channels, such as in packet communication where packets get dropped, it would be beneficial to take the approach of repeated utterances for robust ASR. In this thesis, we have formulated a set of algorithms for both joint evaluation/decoding for recognizing noisy test utterances as well as utilize the same formulation for selective training of Hidden Markov Models (HMMs), again for robust performance. We first address joint recognition of multiple speech patterns given that they belong to the same class. We formulated this problem considering the patterns as isolated words. If there are K test patterns (K ≥ 2) of a word by a speaker, we show that it is possible to improve the speech recognition accuracy over independent single pattern evaluation of test speech, for the case of both clean and noisy speech. We also find the state sequence which best represents the K patterns. This formulation can be extended to connected word recognition or continuous speech recognition also. Next, we consider the benefits of joint multi-pattern likelihood for HMM training. In the usual HMM training, all the training data is utilized to arrive at a best possible parametric model. But, it is possible that the training data is not all genuine and therefore may have labeling errors, noise corruptions, or plain outlier exemplars. Such outliers will result in poorer models and affect speech recognition performance. So it is important to selectively train them so that the outliers get a lesser weightage. Giving lesser weight to an entire outlier pattern has been addressed before in speech recognition literature. However, it is possible that only some portions of a training pattern are corrupted. So it is important that only the corrupted portions of speech are given a lesser weight during HMM training and not the entire pattern. Since in HMM training, multiple patterns of speech from each class are used, we show that it is possible to use joint evaluation methods to selectively train HMMs such that only the corrupted portions of speech are given a lesser weight and not the entire speech pattern. Thus, we have addressed all the three main tasks of a HMM, to jointly utilize the availability of multiple patterns belonging to the same class. We experimented the new algorithms for Isolated Word Recognition in the case of both clean speech and noisy speech. Significant improvement in speech recognition performance is obtained, especially for speech affected by transient/burst noise.

Η παρούσα διατριβή εισάγει μεθόδους για τον αυτόματο τεμαχισμό σημάτων ομιλίας. Συγκεκριμένα παρουσιάζονται τέσσερις νέες μέθοδοι για τον αυτόματο τεμαχισμό σημάτων ομιλίας, τόσο για γλωσσολογικά περιορισμένα όσο και μη προβλήματα. Η πρώτη μέθοδος κάνει χρήση των σημείων του σήματος που αντιστοιχούν στα ανοίγματα των φωνητικών χορδών κατά την διάρκεια της ομιλίας για να εξάγει όρια ψευδό-φωνημάτων με χρήση του αλγορίθμου δυναμικής παραμόρφωσης χρόνου. Η δεύτερη τεχνική εισάγει μια καινοτόμα υβριδική μέθοδο εκπαίδευσης κρυμμένων μοντέλων Μαρκώφ, η οποία τα καθιστά πιο αποτελεσματικά στον τεμαχισμό της ομιλίας. Η τρίτη μέθοδος χρησιμοποιεί αλγορίθμους μαθηματικής παλινδρόμησης για τον συνδυασμό ανεξαρτήτων μηχανών τεμαχισμού ομιλίας. Η τέταρτη μέθοδος εισάγει μια επέκταση του αλγορίθμου Βιτέρμπι με χρήση πολλαπλών παραμετρικών τεχνικών για τον τεμαχισμό της ομιλίας. Τέλος, οι προτεινόμενες μέθοδοι τεμαχισμού χρησιμοποιούνται για την βελτίωση συστημάτων στο πρόβλημα της σύνθεσης ομιλίας, αναγνώρισης ομιλίας και αναγνώρισης γλώσσας. / The present dissertation introduces methods for the automatic segmentation of speech signals. In detail, four new segmentation methods are presented both in for the cases of linguistically constrained or not segmentation. The first method uses pitchmark points to extract pseudo-phonetic boundaries using dynamic time warping algorithm. The second technique introduces a new hybrid method for the training of hidden Markov models, which makes them more effective in the speech segmentation task. The third method uses regression algorithms for the fusion of independent segmentation engines. The fourth method is an extension of the Viterbi algorithm using multiple speech parameterization techniques for segmentation. Finally, the proposed methods are used to improve systems in the task of speech synthesis, speech recognition and language recognition.

Dans les maisons et bâtiments intelligents, il devient nécessaire de limiter l'intervention humaine sur le système énergétique, afin de fluctuer automatiquement l'énergie consommée par les appareils consommateurs. Pour cela, un système de mesure de la consommation électrique d'équipements est aussi nécessaire et peut être déployé de deux façons : intrusive ou non-intrusive. La première solution consiste à relever la consommation de chaque appareil, ce qui est inenvisageable à une grande échelle pour des raisons pratiques liées à l'entretien et aux coûts. Donc, la solution non-intrusive (NILM pour Non-Intrusive Load Monitoring), qui est capable d'identifier les différents appareils en se basant sur les signatures extraites d'une consommation globale, est plus prometteuse. Le problème le plus difficile des algorithmes NILM est comment discriminer les appareils qui ont la même caractéristique énergétique. Pour surmonter ce problème, dans cette thèse, nous proposons d'utiliser une information externe pour améliorer la performance des algorithmes existants. Les premières informations additionnelles proposées considèrent l'état précédent de chaque appareil comme la probabilité de transition d'état ou la distance de Hamming entre l'état courant et l'état précédent. Ces informations sont utilisées pour sélectionner l'ensemble le plus approprié des dispositifs actifs parmi toutes les combinaisons possibles. Nous résolvons ce problème de minimisation en norme l1 par un algorithme d'exploration exhaustive. Nous proposons également d'utiliser une autre information externe qui est la probabilité de fonctionnement de chaque appareil fournie par un réseau de capteurs sans fil (WSN pour Wireless Sensor Network) déployé dans le bâtiment. Ce système baptisé SmartSense, est différent de la solution intrusive car seul un sous-ensemble de tous les dispositifs est surveillé par les capteurs, ce qui rend le système moins intrusif. Trois approches sont appliquées dans le système SmartSense. La première approche applique une détection de changements de niveau sur le signal global de puissance consommé et les compare avec ceux existants pour identifier les dispositifs correspondants. La deuxième approche vise à résoudre le problème de minimisation en norme l1 avec les algorithmes heuristiques de composition Paréto-algébrique et de programmation dynamique. Les résultats de simulation montrent que la performance des algorithmes proposés augmente significativement avec la probabilité d'opération des dispositifs surveillés par le WSN. Comme il n'y a qu'un sous-ensemble de tous les appareils qui sont surveillés par les capteurs, ceux qui sont sélectionnés doivent satisfaire quelques critères tels qu'un taux d'utilisation élevé ou des confusions dans les signatures sélectionnées avec celles des autres. / In smart homes, human intervention in the energy system needs to be eliminated as much as possible and an energy management system is required to automatically fluctuate the power consumption of the electrical devices. To design such system, a load monitoring system is necessary to be deployed in two ways: intrusive or non-intrusive. The intrusive approach requires a high deployment cost and too much technical intervention in the power supply. Therefore, the Non-Intrusive Load Monitoring (NILM) approach, in which the operation of a device can be detected based on the features extracted from the aggregate power consumption, is more promising. The difficulty of any NILM algorithm is the ambiguity among the devices with the same power characteristics. To overcome this challenge, in this thesis, we propose to use an external information to improve the performance of the existing NILM algorithms. The first proposed additional features relate to the previous state of each device such as state transition probability or the Hamming distance between the current state and the previous state. They are used to select the most suitable set of operating devices among all possible combinations when solving the l1-norm minimization problem of NILM by a brute force algorithm. Besides, we also propose to use another external feature that is the operating probability of each device provided by an additional Wireless Sensor Network (WSN). Different from the intrusive load monitoring, in this so-called SmartSense system, only a subset of all devices is monitored by the sensors, which makes the system quite less intrusive. Two approaches are applied in the SmartSense system. The first approach applies an edge detector to detect the step-changes on the power signal and then compare with the existing library to identify the corresponding devices. Meanwhile, the second approach tries to solve the l1-norm minimization problem in NILM with a compositional Pareto-algebraic heuristic and dynamic programming algorithms. The simulation results show that the performance of the proposed algorithms is significantly improved with the operating probability of the monitored devices provided by the WSN. Because only part of the devices are monitored, the selected ones must satisfy some criteria including high using rate and more confusions on the selected patterns with the others.

Tato diplomová práce pojednává o využití algoritmu Dynamic Time Warping (DTW) pro automatické hodnocení výslovnosti anglického jazyka. Práce se zaměřuje na vylepšení již existující aplikace pro výuku výslovnosti, a to ve třech oblastech: uživatelské rozhraní, samotný algoritmus a korektivní zpětná vazba uživateli. První část se věnuje přehledu technik používaných v této oblasti, následně je představen nový design uživatelského rozhraní, popsán navržený systém a experimenty. Experimenty se zaměřují na problematiku detekce chyb na úrovni fonémů, na detekci chyb v primárním důrazu na úrovni slabik a na hodnocení intonace na úrovni slov. Všechny použité metody jsou navrženy tak, aby poskytovaly korektivní zpětnou vazbu uživateli. V poslední části je popsáno, jak byly všechny tři vylepšené oblasti aplikace otestovány.

This work is dedicated to dynamic time warping in biosignal processing, especially it´s application for ECG signals. On the beginning the theoretical notes about cardiography are summarized. Then, the DTW analysis follows along with conditions and demands assessments for it’s successful application. Next, several variants and application possibilities are described. The practical part covers the design of this method, the outputs comprehension, settings optimization and realization of methods related with DTW

This thesis describes the development of voice module, that can recognize simple speech commands by comparation of input sound with recorded templates. The first part of thesis contains a description of used algorithm and a verification of its functionality. The algorithm is based on Mel-frequency cepstral coefficients and dynamic time warping. Thereafter the hardware of voice module is designed, containing signal controller 56F805 from Freescale. The signal from microphone is conditioned by operational amplifiers and digital filter. The third part deals with the development of software for the controller and describes the fixed point implementation of the algorithm, respecting limited capabilities of the controller. Final test proves the usability of voice module in low-noise environment.

Tato diplomová práce se zabývá moderními přístupy detekce klíčových slov a detekce frází v řečových datech. V úvodní části je seznámení s problematikou a teoretický popis metod pro detekci. Následuje popis reprezentace vstupních datových sad použitých při experimentech a evaluaci. Dále jsou uvedeny metody pro detekci klíčových slov definovaných vzorem. Následně jsou popsány evaluační metody a techniky použité pro skórování. Po provedení experimentů na datových sadách a po evaluaci jsou diskutovány výsledky. V dalším kroku jsou navrženy a poté implementovány moderní postupy vedoucí k vylepšení systému pro detekci a opět je provedena evaluace a diskuze dosažených výsledků. V závěrečné části je práce zhodnocena a jsou zde navrženy další směy vývoje našeho systému. Příloha obsahuje manuál pro používání implementovaných skriptů.

The goal of the thesis is being familiar with methods a techniques used in speech processing. Using the obtained knowledge I propose three JavaApplets demonstrating selected methods. In this thesis we can find the theoretical analysis of selected problems.

This dissertation deals with alternative techniques for analysis of genetic information of organisms. The theoretical part presents two different approaches for evaluation of relationship between organisms based on mutual similarity of genetic information contained in their DNA sequences. The first approach is currently standardized phylogenetics analysis of character based records of DNA sequences. Although this approach is computationally expensive due to the need of multiple sequence alignment, it allows evaluation of global and local similarity of DNA sequences. The second approach is represented by techniques for classification of DNA sequences in a form of numerical vectors representing characteristic features of their genetic information. These methods known as „alignment free“ allow fast evaluation of global similarity but cannot evaluate local changes. The new method presented in this dissertation combines the advantages of both approaches. It utilizes numerical representation similar to 1D digital signal, i.e. representation that contains specific trend along x-axis. The experimental part of dissertation deals with design of a set of appropriate tools for genomic signal processing to allow evaluation mutual similarity of taxonomically specific trends. On the basis of the mutual similarity of genomic signals, the classification in the form of dendrogram is applied. It corresponds to phylogenetic trees used in standard phylogenetics.