Fingerprint Recognition: Enhancement, Feature Extraction and Automatic Evaluation of Algorithms

The identification of people by measuring some traits of individual anatomy or physiology has led to a specific research area called biometric recognition. This thesis is focused on improving fingerprint recognition systems considering three important problems: fingerprint enhancement, fingerprint orientation extraction and automatic evaluation of fingerprint algorithms. An effective extraction of salient fingerprint features depends on the quality of the input fingerprint. If the fingerprint is very noisy, we are not able to detect a reliable set of features. A new fingerprint enhancement method, which is both iterative and contextual, is proposed. This approach detects high-quality regions in fingerprints, selectively applies contextual filtering and iteratively expands like wildfire toward low-quality ones. A precise estimation of the orientation field would greatly simplify the estimation of other fingerprint features (singular points, minutiae) and improve the performance of a fingerprint recognition system. The fingerprint orientation extraction is improved following two directions. First, after the introduction of a new taxonomy of fingerprint orientation extraction methods, several variants of baseline methods are implemented and, pointing out the role of pre- and post- processing, we show how to improve the extraction. Second, the introduction of a new hybrid orientation extraction method, which follows an adaptive scheme, allows to improve significantly the orientation extraction in noisy fingerprints. Scientific papers typically propose recognition systems that integrate many modules and therefore an automatic evaluation of fingerprint algorithms is needed to isolate the contributions that determine an actual progress in the state-of-the-art. The lack of a publicly available framework to compare fingerprint orientation extraction algorithms, motivates the introduction of a new benchmark area called FOE (including fingerprints and manually-marked orientation ground-truth) along with fingerprint matching benchmarks in the FVC-onGoing framework. The success of such framework is discussed by providing relevant statistics: more than 1450 algorithms submitted and two international competitions.

Food Safety and Mineral Oil Contaminated Paperboard Packaging: an Analytical Challenge and a Migration Study

Food packaging protects food, but it can sometimes become a source of undesired contaminants. Paper based materials, despite being perceived as “natural” and safe, can contain volatile contaminants (especially if made from recycled paper) able to migrate to food, as mineral oil, phthalates and photoinitiators. Mineral oil is a petroleum product used as printing ink solvent for newspapers, magazines and packaging. From paperboard printing and from recycled fibers (if present), mineral oil migrates into food, even if dry, through the gas phase. Its toxicity is not fully evaluated, but a temporary Acceptable Daily Intake (ADI) of 0.6 mg kg-1 has been established for saturated mineral oil hydrocarbons (MOSH), while aromatic hydrocarbons (MOAH) are more toxic. Extraction and analysis of MOSH and MOAH is difficult due to the thousands of molecules present. Extraction methods for packaging and food have been optimized, then applied for a “shopping trolley survey” on over 100 Italian and Swiss market products. Instrumental analyses were performed with online LC-GC/FID. Average concentration of MOSH in paperboards was 626 mg kg-1. Many had the potential of contaminating foods exceeding temporary ADI tens of times. A long term migration study was then designed to better understand migration kinetics. Egg pasta and müesli were chosen as representative (high surface/weight ratio). They were stored at different temperatures (4, 20, 30, 40 and 60°C) and conditions (free, shelved or boxed packs) for 1 year. MOSH and MOAH kinetic curves show that migration is a fast process, mostly influenced by temperature: in egg pasta (food in direct contact with paperboard), half of MOSH is transferred to food in a week at 40°C and in 8 months at 20°C. The internal plastic bag present in müesli slowed down the startup of migration, creating a “lag time” in the curves.

A new method for the estimation of the reverberation time from measured room impulse responses with application to Italian opera houses

The Schroeder's backward integration method is the most used method to extract the decay curve of an acoustic impulse response and to calculate the reverberation time from this curve. In the literature the limits and the possible improvements of this method are widely discussed. In this work a new method is proposed for the evaluation of the energy decay curve. The new method has been implemented in a Matlab toolbox. Its performance has been tested versus the most accredited literature method. The values of EDT and reverberation time extracted from the energy decay curves calculated with both methods have been compared in terms of the values themselves and in terms of their statistical representativeness. The main case study consists of nine Italian historical theatres in which acoustical measurements were performed. The comparison of the two extraction methods has also been applied to a critical case, i.e. the structural impulse responses of some building elements. The comparison underlines that both methods return a comparable value of the T30. Decreasing the range of evaluation, they reveal increasing differences; in particular, the main differences are in the first part of the decay, where the EDT is evaluated. This is a consequence of the fact that the new method returns a “locally" defined energy decay curve, whereas the Schroeder's method accumulates energy from the tail to the beginning of the impulse response. Another characteristic of the new method for the energy decay extraction curve is its independence on the background noise estimation. Finally, a statistical analysis is performed on the T30 and EDT values calculated from the impulse responses measurements in the Italian historical theatres. The aim of this evaluation is to know whether a subset of measurements could be considered representative for a complete characterization of these opera houses.

Computational methods for genome screening

Motivation An actual issue of great interest, both under a theoretical and an applicative perspective, is the analysis of biological sequences for disclosing the information that they encode. The development of new technologies for genome sequencing in the last years, opened new fundamental problems since huge amounts of biological data still deserve an interpretation. Indeed, the sequencing is only the first step of the genome annotation process that consists in the assignment of biological information to each sequence. Hence given the large amount of available data, in silico methods became useful and necessary in order to extract relevant information from sequences. The availability of data from Genome Projects gave rise to new strategies for tackling the basic problems of computational biology such as the determination of the tridimensional structures of proteins, their biological function and their reciprocal interactions. Results The aim of this work has been the implementation of predictive methods that allow the extraction of information on the properties of genomes and proteins starting from the nucleotide and aminoacidic sequences, by taking advantage of the information provided by the comparison of the genome sequences from different species. In the first part of the work a comprehensive large scale genome comparison of 599 organisms is described. 2,6 million of sequences coming from 551 prokaryotic and 48 eukaryotic genomes were aligned and clustered on the basis of their sequence identity. This procedure led to the identification of classes of proteins that are peculiar to the different groups of organisms. Moreover the adopted similarity threshold produced clusters that are homogeneous on the structural point of view and that can be used for structural annotation of uncharacterized sequences. The second part of the work focuses on the characterization of thermostable proteins and on the development of tools able to predict the thermostability of a protein starting from its sequence. By means of Principal Component Analysis the codon composition of a non redundant database comprising 116 prokaryotic genomes has been analyzed and it has been showed that a cross genomic approach can allow the extraction of common determinants of thermostability at the genome level, leading to an overall accuracy in discriminating thermophilic coding sequences equal to 95%. This result outperform those obtained in previous studies. Moreover, we investigated the effect of multiple mutations on protein thermostability. This issue is of great importance in the field of protein engineering, since thermostable proteins are generally more suitable than their mesostable counterparts in technological applications. A Support Vector Machine based method has been trained to predict if a set of mutations can enhance the thermostability of a given protein sequence. The developed predictor achieves 88% accuracy.

Kernel Methods for Tree Structured Data

Machine learning comprises a series of techniques for automatic extraction of meaningful information from large collections of noisy data. In many real world applications, data is naturally represented in structured form. Since traditional methods in machine learning deal with vectorial information, they require an a priori form of preprocessing. Among all the learning techniques for dealing with structured data, kernel methods are recognized to have a strong theoretical background and to be effective approaches. They do not require an explicit vectorial representation of the data in terms of features, but rely on a measure of similarity between any pair of objects of a domain, the kernel function. Designing fast and good kernel functions is a challenging problem. In the case of tree structured data two issues become relevant: kernel for trees should not be sparse and should be fast to compute. The sparsity problem arises when, given a dataset and a kernel function, most structures of the dataset are completely dissimilar to one another. In those cases the classifier has too few information for making correct predictions on unseen data. In fact, it tends to produce a discriminating function behaving as the nearest neighbour rule. Sparsity is likely to arise for some standard tree kernel functions, such as the subtree and subset tree kernel, when they are applied to datasets with node labels belonging to a large domain. A second drawback of using tree kernels is the time complexity required both in learning and classification phases. Such a complexity can sometimes prevents the kernel application in scenarios involving large amount of data. This thesis proposes three contributions for resolving the above issues of kernel for trees. A first contribution aims at creating kernel functions which adapt to the statistical properties of the dataset, thus reducing its sparsity with respect to traditional tree kernel functions. Specifically, we propose to encode the input trees by an algorithm able to project the data onto a lower dimensional space with the property that similar structures are mapped similarly. By building kernel functions on the lower dimensional representation, we are able to perform inexact matchings between different inputs in the original space. A second contribution is the proposal of a novel kernel function based on the convolution kernel framework. Convolution kernel measures the similarity of two objects in terms of the similarities of their subparts. Most convolution kernels are based on counting the number of shared substructures, partially discarding information about their position in the original structure. The kernel function we propose is, instead, especially focused on this aspect. A third contribution is devoted at reducing the computational burden related to the calculation of a kernel function between a tree and a forest of trees, which is a typical operation in the classification phase and, for some algorithms, also in the learning phase. We propose a general methodology applicable to convolution kernels. Moreover, we show an instantiation of our technique when kernels such as the subtree and subset tree kernels are employed. In those cases, Direct Acyclic Graphs can be used to compactly represent shared substructures in different trees, thus reducing the computational burden and storage requirements.

Development of Original Analytical Methods for the Determination of Drugs of Abuse in Biological Samples

Drug abuse is a major global problem which has a strong impact not only on the single individual but also on the entire society. Among the different strategies that can be used to address this issue an important role is played by identification of abusers and proper medical treatment. This kind of therapy should be carefully monitored in order to discourage improper use of the medication and to tailor the dose according to the specific needs of the patient. Hence, reliable analytical methods are needed to reveal drug intake and to support physicians in the pharmacological management of drug dependence. In the present Ph.D. thesis original analytical methods for the determination of drugs with a potential for abuse and of substances used in the pharmacological treatment of drug addiction are presented. In particular, the work has been focused on the analysis of ketamine, naloxone and long-acting opioids (buprenorphine and methadone), oxycodone, disulfiram and bupropion in human plasma and in dried blood spots. The developed methods are based on the use of high performance liquid chromatography (HPLC) coupled to various kinds of detectors (mass spectrometer, coulometric detector, diode array detector). For biological sample pre-treatment different techniques have been exploited, namely solid phase extraction and microextraction by packed sorbent. All the presented methods have been validated according to official guidelines with good results and some of these have been successfully applied to the therapeutic drug monitoring of patients under treatment for drug abuse.

Local Trigonometric Methods for Time Series Smoothing.

The thesis is concerned with local trigonometric regression methods. The aim was to develop a method for extraction of cyclical components in time series. The main results of the thesis are the following. First, a generalization of the filter proposed by Christiano and Fitzgerald is furnished for the smoothing of ARIMA(p,d,q) process. Second, a local trigonometric filter is built, with its statistical properties. Third, they are discussed the convergence properties of trigonometric estimators, and the problem of choosing the order of the model. A large scale simulation experiment has been designed in order to assess the performance of the proposed models and methods. The results show that local trigonometric regression may be a useful tool for periodic time series analysis.

Knowledge Patterns for the Web: extraction, tranformation and reuse

This thesis aims at investigating methods and software architectures for discovering what are the typical and frequently occurring structures used for organizing knowledge in the Web. We identify these structures as Knowledge Patterns (KPs). KP discovery needs to address two main research problems: the heterogeneity of sources, formats and semantics in the Web (i.e., the knowledge soup problem) and the difficulty to draw relevant boundary around data that allows to capture the meaningful knowledge with respect to a certain context (i.e., the knowledge boundary problem). Hence, we introduce two methods that provide different solutions to these two problems by tackling KP discovery from two different perspectives: (i) the transformation of KP-like artifacts to KPs formalized as OWL2 ontologies; (ii) the bottom-up extraction of KPs by analyzing how data are organized in Linked Data. The two methods address the knowledge soup and boundary problems in different ways. The first method provides a solution to the two aforementioned problems that is based on a purely syntactic transformation step of the original source to RDF followed by a refactoring step whose aim is to add semantics to RDF by select meaningful RDF triples. The second method allows to draw boundaries around RDF in Linked Data by analyzing type paths. A type path is a possible route through an RDF that takes into account the types associated to the nodes of a path. Then we present K~ore, a software architecture conceived to be the basis for developing KP discovery systems and designed according to two software architectural styles, i.e, the Component-based and REST. Finally we provide an example of reuse of KP based on Aemoo, an exploratory search tool which exploits KPs for performing entity summarization.

Secondary Microseisms Characterization and Green’s Function Extraction at the Larderello-Travale Geothermal Field (Italy)

Over the past ten years, the cross-correlation of long-time series of ambient seismic noise (ASN) has been widely adopted to extract the surface-wave part of the Green’s Functions (GF). This stochastic procedure relies on the assumption that ASN wave-field is diffuse and stationary. At frequencies <1Hz, the ASN is mainly composed by surface-waves, whose origin is attributed to the sea-wave climate. Consequently, marked directional properties may be observed, which call for accurate investigation about location and temporal evolution of the ASN-sources before attempting any GF retrieval. Within this general context, this thesis is aimed at a thorough investigation about feasibility and robustness of the noise-based methods toward the imaging of complex geological structures at the local (∼10-50km) scale. The study focused on the analysis of an extended (11 months) seismological data set collected at the Larderello-Travale geothermal field (Italy), an area for which the underground geological structures are well-constrained thanks to decades of geothermal exploration. Focusing on the secondary microseism band (SM;f>0.1Hz), I first investigate the spectral features and the kinematic properties of the noise wavefield using beamforming analysis, highlighting a marked variability with time and frequency. For the 0.1-0.3Hz frequency band and during Spring- Summer-time, the SMs waves propagate with high apparent velocities and from well-defined directions, likely associated with ocean-storms in the south- ern hemisphere. Conversely, at frequencies >0.3Hz the distribution of back- azimuths is more scattered, thus indicating that this frequency-band is the most appropriate for the application of stochastic techniques. For this latter frequency interval, I tested two correlation-based methods, acting in the time (NCF) and frequency (modified-SPAC) domains, respectively yielding esti- mates of the group- and phase-velocity dispersions. Velocity data provided by the two methods are markedly discordant; comparison with independent geological and geophysical constraints suggests that NCF results are more robust and reliable.

Original analytical methods for the determination of psychoactive compounds in complex matrices

This thesis work aims to develop original analytical methods for the determination of drugs with a potential for abuse, for the analysis of substances used in the pharmacological treatment of drug addiction in biological samples and for the monitoring of potentially toxic compounds added to street drugs. In fact reliable analytical techniques can play an important role in this setting. They can be employed to reveal drug intake, allowing the identification of drug users and to assess drug blood levels, assisting physicians in the management of the treatment. Pharmacological therapy needs to be carefully monitored indeed in order to optimize the dose scheduling according to the specific needs of the patient and to discourage improper use of the medication. In particular, different methods have been developed for the detection of gamma-hydroxybutiric acid (GHB), prescribed for the treatment of alcohol addiction, of glucocorticoids, one of the most abused pharmaceutical class to enhance sport performance and of adulterants, pharmacologically active compounds added to illicit drugs for recreational purposes. All the presented methods are based on capillary electrophoresis (CE) and high performance liquid chromatography (HPLC) coupled to various detectors (diode array detector, mass spectrometer). Biological samples pre-treatment was carried out using different extraction techniques, liquid-liquid extraction (LLE) and solid phase extraction (SPE). Different matrices have been considered: human plasma, dried blood spots, human urine, simulated street drugs. These developed analytical methods are individually described and discussed in this thesis work.