We present a component-based approach for recognizing objectsunder large pose changes. From a set of training images of a givenobject we extract a large number of components which are clusteredbased on the similarity of their image features and their locations withinthe object image. The cluster centers build an initial set of componenttemplates from which we select a subset for the final recognizer.In experiments we evaluate different sizes and types of components andthree standard techniques for component selection. The component classifiersare finally compared to global classifiers on a database of fourobjects.

Η παρούσα διπλωματική εργασία αναφέρεται σε ένα αυτόματο σύστημα αναγνώρισης με χρήση της ομιλίας, και πιο συγκεκριμένα σε ένα σύστημα αναγνώρισης φύλου μέσω ομιλίας. Αναλύεται η δομή του, περιγράφεται η λειτουργία του και δίνονται οι λεπτομέρειες κάθε τμήματος του. Αρχικά, η εργασία επικεντρώνεται στην προεπεξεργασία του σήματος ομιλίας και στην εξαγωγή των κατάλληλων παραμέτρων, οι οποίες θα μπορέσουν να χαρακτηρίσουν κάθε φύλο. Στη συνέχεια, περιγράφεται η διαδικασία ταξινόμησης του συστήματος, οι αλγόριθμοι που χρησιμοποιούνται και στο τέλος παρουσιάζονται τα ποσοστά επιτυχίας. Τα αποτελέσματα υποδεικνύουν και το βέλτιστο σύνολο παραμέτρων ομιλίας για αξιόπιστη αναγνώριση φύλου. / The purpose of this diploma thesis is the study of a gender recognition system based on speech. More specifically the system’s structure is analyzed, its functions are described and details regarding every single part are given. We focus on the preprocessing of the speech signal and the definition of the appropriate parameters that characterize every gender. Moreover, the methods, which are used for classification during the experimental setup, are described and be presented with their results. These results also suggest the optimized speech parameters appropriate for reliable gender recognition.

Many real-world face recognition tasks are under unconstrained conditions such as off-angle pose variations, illumination variations, facial occlusion, facial expression, etc. In this work, we are focusing on the real-world scenarios where only the periocular region of a face is visible such as in the ISIS case. In Part I of the dissertation, we will showcase the face recognition capability based on the periocular region, which we call the periocular face recognition. We will demonstrate that face matching using the periocular region directly is more robust than the full face in terms of age-tolerant face recognition, expression-tolerant face recognition, pose-tolerant face recognition, as well as contains more cues for determining the gender information of a subject. In this dissertation, we will study direct periocular matching more comprehensively and systematically using both shallow and deep learning methods. Based on this, in Part II and Part III of the dissertation, we will continue to explore an indirect way of carrying out the periocular face recognition: periocular-based full face hallucination, because we want to capitalize on the powerful commercial face matchers and deep learning-based face recognition engines which are all trained on large-scale full face images. The reproducibility and feasibility of re-training for a proprietary facial region, such as the periocular region, is relatively low, due to the nonopen source nature of commercial face matchers as well as the amount of training data and computation power required by the deep learning based models. We will carry out the periocular-based full face hallucination based on two proposed reconstructive dictionary learning methods, including the dimensionally weighted K-SVD (DW-KSVD) dictionary learning approach and its kernel feature space counterpart using Fastfood kernel expansion approximation to reconstruct high-fidelity full face images from the periocular region, as well as two proposed generative deep learning approaches that build upon deep convolutional generative adversarial networks (DCGAN) to generate the full face from the periocular region observations, including the Gang of GANs (GoGAN) method and the discriminant nonlinear many-to-one generative adversarial networks (DNMM-GAN) for applications such as the generative open-set landmark-free frontalization (Golf) for faces and universal face optimization (UFO), which tackles an even broader set of problems than periocular based full face hallucination. Throughout Parts I-III, we will study how to handle challenging realworld scenarios such as unconstrained pose variations, unconstrained illumination conditions, and unconstrained low resolution of the periocular and facial images. Together, we aim to achieve unconstrained periocular face recognition through both direct periocular face matching and indirect periocular-based full face hallucination. In the final Part IV of the dissertation, we will go beyond and explore several new methods in deep learning that are statistically efficient for generalpurpose image recognition. Methods include the local binary convolutional neural networks (LBCNN), the perturbative neural networks (PNN), and the polynomial convolutional neural networks (PolyCNN).

Speech in a noisy background presents a challenge for the recognition of that speech both by human listeners and by computers tasked with understanding human speech (automatic speech recognition; ASR). Years of research have resulted in many solutions, though none so far have completely solved the problem. Current solutions generally require some form of estimation of the noise, in order to remove it from the signal. The limitation is that noise can be highly unpredictable and highly variable, both in form and loudness. The present report proposes a method of recording a speech signal in a noisy environment that largely prevents noise from reaching the recording microphone. This method utilizes the human skull as a noise-attenuation device by placing the microphone in the ear canal. For further noise dampening, a pair of noise-reduction earmuffs are used over the speakers' ears. A corpus of speech was recorded with a microphone in the ear canal, while also simultaneously recording speech at the mouth. Noise was emitted from a loudspeaker in the background. Following the data collection, the speech recorded at the ear was analyzed. A substantial noise-reduction benefit was found over mouth-recorded speech. However, this speech was missing much high-frequency information. With minor processing, mid-range frequencies were amplified, increasing the intelligibility of the speech. A human perception task was conducted using both the ear-recorded and mouth-recorded speech. Participants in this experiment were significantly more likely to understand ear-recorded speech over the noisy, mouth-recorded speech. Yet, participants found mouth-recorded speech with no noise the easiest to understand. These recordings were also used with an ASR system. Since the ear-recorded speech is missing much high-frequency information, it did not recognize the ear-recorded speech readily. However, when an acoustic model was trained low-pass filtered speech, performance improved. These experiments demonstrated that humans, and likely an ASR system, with additional training, would be able to more easily recognize ear-recorded speech than speech in noise. Further speech processing and training may be able to improve the signal's intelligibility for both human and automatic speech recognition.

Introduction: Impairments in facial affect recognition have been linked to the development of various disorders. The aim of the current work is to conduct a systematic review and meta-analysis of studies examining whether this ability is impaired in males with psychopathy or antisocial traits, when compared to healthy individuals. Method: Studies were eligible for inclusion if they compared facial affect recognition in either a) psychopathic vs. antisocial males, b) psychopathic vs. healthy controls and c) antisocial vs. healthy controls. Primary outcomes were group differences in overall emotion recognition, fear recognition, and sadness recognition. Secondary outcomes were differences in recognition of disgust, happiness, surprise and anger. Results: Fifteen papers comprising 214 psychopathic males, 491 antisocial males and 386 healthy community controls were identified. In psychopathy, limited evidence suggested impairments in fear (k=2), sadness (k=1) and surprise (k=1) recognition relative to healthy individuals, but overall affect recognition ability was not affected (k=2). Findings were inconclusive for antisocial (k=4-6), although impairments in surprise (k=4) and disgust (k=5) recognition were observed. Psychopathic and antisocial samples did not differ in their ability to detect sadness (k=4), but psychopaths were less able to recognise happiness (k=4) and surprise (k=3). Conclusion: Limited evidence suggests psychopathic and antisocial personality traits are associated with small to moderate deficits in specific aspects of emotion recognition. However considerable heterogeneity was identified, and study quality was often poor. Adequately powered studies using validated assessment measures, rater masking and a priori public registration of hypotheses and methods are required.

>Magister Scientiae - MSc / Hand shape recognition is a pivotal part of any system that attempts to implement Sign Language recognition. This thesis presents a novel system which recognises hand shapes from a single camera view in 2D. By mapping the recognised hand shape from 2D to 3D,it is possible to obtain 3D co-ordinates for each of the joints within the hand using the kinematics embedded in a 3D hand avatar and smooth the transformation in 3D space between any given hand shapes. The novelty in this system is that it does not require a hand pose to be recognised at every frame, but rather that hand shapes be detected at a given step size. This architecture allows for a more efficient system with better accuracy than other related systems. Moreover, a real-time hand tracking strategy was developed that works efficiently for any skin tone and a complex background.

Identication of cartridge is very essential in the field of forensics, military or people who collect ammunitions. The cartridges can beidentied by their headstamps.This thesis presents work on identification and matching of cartridge headstamp from the image. The Libor Masek's open source iris recognition algorithm is considered for the identification of cartridge pattern from the image.The dataset is devoleped with the cartridge headstamp patterns and matching of cartridge headstamp patterns is implemented. For matching of the cartridge pattern the Hamming distance is considered as the metric to differentiate interclass and intraclass comparisons. Variance is used as a criteria to discard the unwanted areas of the cartridge headstamp pattern.Four distinct cartridge headstamp patterns are considered. Three cartridges of each headstamp pattern are considered for intra class comparisons. The validation of the method is performed.

No description available.

El texto completo de este trabajo no está disponible en el Repositorio Académico UPC por restricciones de la casa editorial donde ha sido publicado. / Peruvian Sign Language Recognition (PSL) is approached as a classification problem. Previous work has employed 2D features from the position of hands to tackle this problem. In this paper, we propose a method to construct a dataset consisting of 3D spatial positions of static gestures from the PSL alphabet, using the HTC Vive device and a well-known technique to extract 21 keypoints from the hand to obtain a feature vector. A dataset of 35, 400 instances of gestures for PSL was constructed and a novel way to extract data was stated. To validate the appropriateness of this dataset, a comparison of four baselines classifiers in the Peruvian Sign Language Recognition (PSLR) task was stated, achieving 99.32% in the average in terms of F1 measure in the best case. / Revisión por pares

The ability to infer and mediate intentions has been recognized as a crucial task in recent robotics research, where it is agreed that robots are required to be equipped with intentional mechanisms in order to participate in collaborative tasks with humans. Reasoning about - or rather, perceiving - intentions enables robots to infer what other agents are doing, to communicate what are their plans, or to take proactive decisions. Intent recognition relates to several system requirements, such as the need of an enhanced collaboration mechanism in human-machine interactions, the need for adversarial technology in competitive scenarios, ambient intelligence, or predictive security systems. When attempting to describe what an intention is, agreement exists to represent it as a plan together with the goal it attempts to achieve. Being compatible with computer science concepts, this representation enables to handle intentions with methodologies based on planning, such as the Planning Domain Description Language or Hierarchical Task Networks. In this licentiate we describe how intentions can be processed using classical planning methods, with an eye also on newer technologies such as deep networks. Our goal is to study and define computational models that would allow robotic agents to infer, construct and mediate intentions. Additionally, we explore how intentions in the form of abstract plans can be grounded to sensorial data, and in particular we provide discussion on grounding over speech utterances and affordances, that correspond to the action possibilities offered by an environment.