1,721,227 research outputs found
Variants of the Borda count method for combining ranked classifier hypotheses
No full textThe Borda count is a simple yet effective method of combining rankings. In pattern recognition, classifiers are often able to return a ranked set of results. Several experiments have been conducted to test the ability of the Borda count and two variant methods to combine these ranked classifier results. By using artificial data, domain-specific results were avoided. The results show the strength of the Borda count when many errors occur in the results, but also show its weakness in case of a limited number of large ranking errors
Variants of the Borda count method for combining ranked classifier hypotheses
No full textThe Borda count is a simple yet effective method of combining rankings. In pattern recognition, classifiers are often able to return a ranked set of results. Several experiments have been conducted to test the ability of the Borda count and two variant methods to combine these ranked classifier results. By using artificial data, domain-specific results were avoided. The results show the strength of the Borda count when many errors occur in the results, but also show its weakness in case of a limited number of large ranking errors
Variants of the Borda count method for combining ranked classifier hypotheses
No full textThe Borda count is a simple yet effective method of combining rankings. In pattern recognition, classifiers are often able to return a ranked set of results. Several experiments have been conducted to test the ability of the Borda count and two variant methods to combine these ranked classifier results. By using artificial data, domain-specific results were avoided. The results show the strength of the Borda count when many errors occur in the results, but also show its weakness in case of a limited number of large ranking errors
Statistical pattern recognition for automatic writer identification and verification
Full text linkThe thesis addresses the problem of automatic person identification using scanned images of handwriting.Identifying the author of a handwritten sample using automatic image-based methods is an interesting pattern recognition problem with direct applicability in the forensic and historic document analysis fields.We describes a number of new and very effective techniques that we have developed in recent years for automatic writer identification and verification.
Writer individuality is encoded using probability distribution functions (PDFs) extracted handwritten text blocks. Our methods operate at two levels of analysis: the texture level and the character-shape (allograph) level. At the texture level, we use contour-based joint directional PDFs that encode orientation and curvature information to give an intimate characterization of the individual handwriting style. In our analysis at the allograph level, the writer is considered to be characterized by a stochastic pattern generator of ink-trace fragments, or graphemes. The PDF of these simple shapes in a given handwriting sample is characteristic for the writer and is computed using a common codebook of graphemes obtained by clustering.
There are two distinguishing characteristics of our approach: human intervention is minimized in the writer identification process and we encode individual handwriting style using features designed to be independent of the textual content of the handwritten sample. In our methods the computer is completely unaware of what has been written in the sample. The handwriting is merely seen as a texture characterized by some directional probability distributions or as a simple stochastic shape-emission process characterized by a grapheme occurrence probability. Our methods were statistically evaluated using datasets with handwriting samples collected from up to 900 subjects.
The development of our writer identification techniques takes place at a time when many biometric methods (e.g. iris, fingerprint, face) undergo a transition from research to real full-scale deployment. Our methods also have practical feasibility and hold the promise of direct applicability.
The main chapters of the thesis are based on a number of scientific papers that were published during the duration of the PhD project. Here is an outline with short abstracts for the main chapters and references to the corresponding papers.
Beyond OCR: Handwritten manuscript attribute understanding
Full text linkKnowing the author, date and location of handwritten historical documents is very important for historians to completely understand and reveal the valuable information they contain. In this thesis, three attributes, such as writer, date and geographical location, are studied by analyzing the handwriting style contained in manuscript images and develop novel algorithms to estimate these attributes on the basis of pattern recognition methods. Handwriting styles are different between different individuals and implicitly encoded in the handwritten patterns when they were written down. This information can be used for writer identification. In this thesis, different features, such as textural-based, textural-free and grapheme-based features, are designed and extracted to present the handwriting style of historical handwritten documents in particular. These features are computational efficient and explainable to end users. According to paleographical expertise, handwriting styles change gradually, continuously and in general within a relatively limited time frame, within 25 years. Modeling the gradual style evolution can be used to date and localize historical manuscripts. This thesis designed a system to date the charters produced between 1300 and 1550 CE in the Medieval Dutch language area. We have shown that designed shape features can be applied quickly and conveniently, without much training efforts on new data sets and problems, even in conditions where the amount of labeled data is relatively limited
Visual attention and active vision : from natural to artificial systems
Full text linkThis dissertation presents a multi-disciplinary study to active vision and visual attention. The topics have been discussed in natural systems in Part I and in artificial systems in Part II. The main theme, the role of symmetry in visual attention, has been studied from both perspectives, which has led to a better understanding of the topic. This chapter first gives an overview of the main results and conclusions discussed in the thesis. This is followed by a discussion dealing with the benefits of a multi-disciplinary approach, the role of symmetry in bottom-up object detection, and the use of other Gestalt principles in saliency models and computer-vision systems....
Zie: Chapter 10
Robust and applicable handwriting biometrics
Full text linkHandschrift bevat kenmerken die kunnen verraden wie de schrijver is, Dat komt van pas wanneer de politie wil achterhalen wie de schrijver is van bijvoorbeeld een handgeschreven dreigbrief of een mogelijk valse zelfmoordbrief. Het is ook nuttig voor geschiedenisonderzoek. want veel historische bronnen zijn handgeschreven en soms is het mogeJijk om iets te leren over onze geschiedenis door te achterhalen welke documenten door dezelfde persoon geschreven moeten zijn.
Het vergelijken van handschrift geheurt traditioneel nog met de hand, door experts, op basis van hun kennis en ervaring. Hoewel zij doorgaans zeer kundig zijn zitten hier wel twee
nadelen aanl: deze aanpak is niet objectief en het is tijclrovend werk. Deze tekortkomingen kunnen goed worden gecompenseerd door de computer. want die heeft tegengcstelde kwaliteiten: algoritmes kunnen onbevooroordeeld en razendsnel een enorne hoeveelheid metingen verricbten. Het ligt dus voor de hand om de traditionele handschriftvergelijking te verrijken met een slim stuk gereedschap in de vorm van een computerprogramma dat handschrift kan vergelijken en daa uitsparken over kan denl.
Zulke' programmatuur noemen we systemen voor handschriftbiometrie. Er zijn twee typen: schrijververificatie en schrijveridentifiatie. Een systeem voor schrijververificatie bepaalt of twee documenten van dezelfde hand zijn; dit kan worden gebruikt om de uitspraak van handschriftdeskundigen te onderbouwen of juist te ontkrachten. Een systeem yoor schrijveridentificatie zoekt op basis van een document in een bestaande collectie documenten naar documenten met vergelijkbaar handschrift en levert daarbij de identileit van de schrijvers daarvan. Dit kan worden gebruikt door de politie om mogelijke daders te vinden, of door historici om documen te groeperen die mogelijk van dezelfde hand zijn. In hoofdstuk 1 wordt de basis van zulke systemen voor voor handschriftbiometrie toegelicht.
Bestaande systemen presteren prima op handschrift dat is geschreven in laboratoriumcondities, maar er is weinig bekend over de prestaties in realistische omstandigheden. In praktijk kan handschrift moeilijkhedenl bevatten zoals een
tekort aan tekst. doorgestreepte woorden en verdraaid handsrhrift.
Going Beyond Counting First Authors in Author Co-citation Analysis
Full text linkThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
Signal-driven sound processing for uncontrolled environments
Full text linkToepassingen van automatische geluidsbronherkenning zijn steeds vaker in het veld in plaats van in het laboratorium. Dat betekent dat een aantal aannames van bestaande technieken niet meer geldig zijn. Met name de aanname dat er maar één bron is kan in het veld niet worden gegarandeerd. In dit onderzoek heb ik een methode ontwikkeld die deze aannames niet maakt. Deze methode kijkt naar kleine omgevingen rond een tijd-frequentie punt en bepaalt of die omgevingen op tonen of pulsen lijken. Omdat het onwaarschijnlijk is dat meerdere bronnen op hetzelfde moment dezelfde frequentie produceren met vergelijkbare energie is het waarschijnlijk dat zo'n toon of puls van één bron afkomstig is. Aansluitende punten met hetzelfde type bron worden gegroepeerd en deze groepen vormen de basis van de automatische geluidsbronherkenning. De herkenning van de groepen gebeurt door een aantal eigenschappen van de groepen vast te stellen. Door de eigenschappen te vergelijken met de eigenschappen van eerder geziene voorbeelden worden de groepen geclassificeerd. De classificatie is dus afhankelijk van het type geluid, toon of puls, en van de eigenschappen van het geluid dat binnenkomt. De methode is getest op een aantal nieuwe datasets, die zijn opgenomen zonder de omgeving te controleren. Een probleem bij dit soort datasets is echter om precies te weten welke geluiden er in voorkomen. Daarom zijn annotaties van geluidsbronnen gemaakt door menselijke luisteraars, die het niet altijd eens blijken te zijn. De kwaliteit van de voorgestelde methode voor automatische herkenning ligt in de buurt van de menselijke annotators.
- …
