Short-term Feature Space and Music Genre Classification

In music genre classification, most approaches rely on statistical characteristics of low-level features computed on short audio frames. In these methods, it is implicitly considered that frames carry equally relevant information loads and that either individual frames, or distributions thereof, somehow capture the specificities of each genre. In this paper we study the representation space defined by short-term audio features with respect to class boundaries, and compare different processing techniques to partition this space. These partitions are evaluated in terms of accuracy on two genre classification tasks, with several types of classifiers. Experiments show that a randomized and unsupervised partition of the space, used in conjunction with a Markov Model classifier lead to accuracies comparable to the state of the art. We also show that unsupervised partitions of the space tend to create less hubs.

Knowledge extraction from medium voltage load diagrams to support the definition of electrical tariffs

With the electricity market liberalization, distribution and retail companies are looking for better market strategies based on adequate information upon the consumption patterns of its electricity customers. In this environment all consumers are free to choose their electricity supplier. A fair insight on the customer´s behaviour will permit the definition of specific contract aspects based on the different consumption patterns. In this paper Data Mining (DM) techniques are applied to electricity consumption data from a utility client’s database. To form the different customer´s classes, and find a set of representative consumption patterns, we have used the Two-Step algorithm which is a hierarchical clustering algorithm. Each consumer class will be represented by its load profile resulting from the clustering operation. Next, to characterize each consumer class a classification model will be constructed with the C5.0 classification algorithm.

Supercritical carbon dioxide extraction of bioactive compounds from microalgae and volatile oils from aromatic plants

A discussion of the most interesting results obtained in our laboratories, during the supercritical CO(2) extraction of bioactive compounds from microalgae and volatile oils from aromatic plants, was carried out. Concerning the microalgae, the studies on Botryococcus braunii and Chlorella vulgaris were selected. Hydrocarbons from the first microalgae, which are mainly linear alkadienes (C(23)-C(31)) with an odd number of carbon atoms, were selectively extracted at 313 K increasing the pressure up to 30.0 MPa. These hydrocarbons are easily extracted at this pressure, since they are located outside the cellular walls. The extraction of carotenoids, mainly canthaxanthin and astaxanthin, from C. vulgaris is more difficult. The extraction yield of these components at 313 K and 35.0 MPa increased with the degree of crushing of the microalga, since they are not extracellular. On the other hand, for the extraction of volatile oils from aromatic plants, studies on Mentha pulegium and Satureja montana L were chosen. For the first aromatic plant, the composition of the volatile and essential oils was similar, the main components being the pulegone and menthone. However, this volatile oil contained small amounts of waxes, which content decreased with decreasing particle size of the plant matrix. For S. montana L it was also observed that both oils have a similar composition, the main components being carvacrol and thymol. The main difference is the relative amount of thymoquinone, which content can be 15 times higher in volatile oil. This oxygenated monoterpene has important biological activities. Moreover, experimental studies on anticholinesterase activity of supercritical extracts of S. montana were also carried out. The supercritical nonvolatile fraction, which presented the highest content of the protocatechuic, vanilic, chlorogenic and (+)-catechin acids, is the most promising inhibitor of the enzyme butyrylcholinesterase. In contrast, the Soxhlet acetone extract did not affect the activity of this enzyme at the concentrations tested. (C) 2011 Elsevier B.V. All rights reserved.

Symbolic Knowledge Extraction from Trained Neural Networks Governed by Lukasiewicz Logics

This work describes a methodology to extract symbolic rules from trained neural networks. In our approach, patterns on the network are codified using formulas on a Lukasiewicz logic. For this we take advantage of the fact that every connective in this multi-valued logic can be evaluated by a neuron in an artificial network having, by activation function the identity truncated to zero and one. This fact simplifies symbolic rule extraction and allows the easy injection of formulas into a network architecture. We trained this type of neural network using a back-propagation algorithm based on Levenderg-Marquardt algorithm, where in each learning iteration, we restricted the knowledge dissemination in the network structure. This makes the descriptive power of produced neural networks similar to the descriptive power of Lukasiewicz logic language, minimizing the information loss on the translation between connectionist and symbolic structures. To avoid redundance on the generated network, the method simplifies them in a pruning phase, using the "Optimal Brain Surgeon" algorithm. We tested this method on the task of finding the formula used on the generation of a given truth table. For real data tests, we selected the Mushrooms data set, available on the UCI Machine Learning Repository.

Lipid content of frozen fish: comparison of different extraction methods and variability during freezing storage

In this work, a microwave-assisted extraction (MAE) methodology was compared with several conventional extraction methods (Soxhlet, Bligh & Dyer, modified Bligh & Dyer, Folch, modified Folch, Hara & Radin, Roese-Gottlieb) for quantification of total lipid content of three fish species: horse mackerel (Trachurus trachurus), chub mackerel (Scomber japonicus), and sardine (Sardina pilchardus). The influence of species, extraction method and frozen storage time (varying from fresh to 9 months of freezing) on total lipid content was analysed in detail. The efficiencies of methods MAE, Bligh & Dyer, Folch, modified Folch and Hara & Radin were the highest and although they were not statistically different, differences existed in terms of variability, with MAE showing the highest repeatability (CV = 0.034). Roese-Gottlieb, Soxhlet, and modified Bligh & Dyer methods were very poor in terms of efficiency as well as repeatability (CV between 0.13 and 0.18).

A novel application of microwave-assisted extraction of polyphenols from brewer’s spent grain with HPLC-DAD-MS analysis

This paper reports a novel application of microwave-assisted extraction (MAE) of polyphenols from brewer’s spent grains (BSG). A 24 orthogonal composite design was used to obtain the optimal conditions of MAE. The influence of the MAE operational parameters (extraction time, temperature, solvent volume and stirring speed) on the extraction yield of ferulic acid was investigated through response surface methodology. The results showed that the optimal conditions were 15 min extraction time, 100 °C extraction temperature, 20 mL of solvent, and maximum stirring speed. Under these conditions, the yield of ferulic acid was 1.31±0.04% (w/w), which was fivefold higher than that obtained with conventional solid–liquid extraction techniques. The developed new extraction method considerably reduces extraction time, energy and solvent consumption, while generating fewer wastes. HPLC-DADMS analysis indicated that other hydroxycinnamic acids and several ferulic acid dehydrodimers, as well as one dehydrotrimer were also present, confirming that BSG is a valuable source of antioxidant compounds.

Remediation of sandy soils contaminated with hydrocarbons and halogenated hydrocarbons by soil vapour extraction

This paper presents the study of the remediation of sandy soils containing six of the most common contaminants (benzene, toluene, ethylbenzene, xylene, trichloroethylene and perchloroethylene) using soil vapour extraction (SVE). The influence of soil water content on the process efficiency was evaluated considering the soil type and the contaminant. For artificially contaminated soils with negligible clay contents and natural organic matter it was concluded that: (i) all the remediation processes presented efficiencies above 92%; (ii) an increase of the soil water content led to a more time-consuming remediation; (iii) longer remediation periods were observed for contaminants with lower vapour pressures and lower water solubilities due to mass transfer limitations. Based on these results an easy and relatively fast procedure was developed for the prediction of the remediation times of real soils; 83% of the remediation times were predicted with relative deviations below 14%.

Sequential application of soil vapor extraction and bioremediation processes for the remediation of ethylbenzene-contaminated soils

Soil vapor extraction (SVE) is an efficient, well-known and widely applied soil remediation technology. However, under certain conditions it cannot achieve the defined cleanup goals, requiring further treatment, for example, through bioremediation (BR). The sequential application of these technologies is presented as a valid option but is not yet entirely studied. This work presents the study of the remediation of ethylbenzene (EB)-contaminated soils, with different soil water and natural organic matter (NOMC) contents, using sequential SVE and BR. The obtained results allow the conclusion that: (1) SVE was sufficient to reach the cleanup goals in 63% of the experiments (all the soils with NOMC below 4%), (2) higher NOMCs led to longer SVE remediation times, (3) BR showed to be a possible and cost-effective option when EB concentrations were lower than 335 mg kgsoil −1, and (4) concentrations of EB above 438 mg kgsoil −1 showed to be inhibitory for microbial activity.

Analysis of polycyclic aromatic hydrocarbons in fish: optimisation and validation of microwave-assisted extraction

An accurate and sensitive method for determination of 18 polycyclic aromatic hydrocarbons (PAHs) (16 PAHs considered by USEPA as priority pollutants, dibenzo[a,l]pyrene and benzo[j]fluoranthene) in fish samples was validated. Analysis was performed by microwave-assisted extraction and liquid chromatography with photodiode array and fluorescence detection. Response surface methodology was used to find the optimal extraction parameters. Validation of the overall methodology was performed by spiking assays at four levels and using SRM 2977. Quantification limits ranging from 0.15–27.16 ng/g wet weight were obtained. The established method was applied in edible tissues of three commonly consumed and commercially valuable fish species (sardine, chub mackerel and horse mackerel) originated from Atlantic Ocean. Variable levels of naphthalene (1.03–2.95 ng/g wet weight), fluorene (0.34–1.09 ng/g wet weight) and phenanthrene (0.34–3.54 ng/g wet weight) were detected in the analysed samples. None of the samples contained detectable amounts of benzo[a]pyrene, the marker used for evaluating the occurrence and carcinogenic effects of PAHs in food.

Extraction of ochratoxin A in bread samples by the QuEChERS methodology

A QuEChERS method for the extraction of ochratoxin A (OTA) from bread samples was evaluated. A factorial design (23) was used to find the optimal QuEChERS parameters (extraction time, extraction solvent volume and sample mass). Extracts were analysed by LC with fluorescence detection. The optimal extraction conditions were: 5 g of sample, 15 mL of acetonitrile and 3 min of agitation. The extraction procedure was validated by systematic recovery experiments at three levels. The recoveries obtained ranged from 94.8% (at 1.0 μg kg -1) to 96.6% (at 3.0 μg kg -1). The limit of quantification of the method was 0.05 μg kg -1. The optimised procedure was applied to 20 samples of different bread types (‘‘Carcaça’’, ‘‘Broa de Milho’’, and ‘‘Broa de Avintes’’) highly consumed in Portugal. None of the samples exceeded the established European legal limit of 3 μg kg -1.

The usefulness of ultrasound in the classification of chronic liver disease

Chronic Liver Disease is a progressive, most of the time asymptomatic, and potentially fatal disease. In this paper, a semi-automatic procedure to stage this disease is proposed based on ultrasound liver images, clinical and laboratorial data. In the core of the algorithm two classifiers are used: a k nearest neighbor and a Support Vector Machine, with different kernels. The classifiers were trained with the proposed multi-modal feature set and the results obtained were compared with the laboratorial and clinical feature set. The results showed that using ultrasound based features, in association with laboratorial and clinical features, improve the classification accuracy. The support vector machine, polynomial kernel, outperformed the others classifiers in every class studied. For the Normal class we achieved 100% accuracy, for the chronic hepatitis with cirrhosis 73.08%, for compensated cirrhosis 59.26% and for decompensated cirrhosis 91.67%.

Comparison of different breast planning techniques and algorithms for radiation therapy treatment

This work aims at investigating the impact of treating breast cancer using different radiation therapy (RT) techniques – forwardly-planned intensity-modulated, f-IMRT, inversely-planned IMRT and dynamic conformal arc (DCART) RT – and their effects on the whole-breast irradiation and in the undesirable irradiation of the surrounding healthy tissues. Two algorithms of iPlan BrainLAB treatment planning system were compared: Pencil Beam Convolution (PBC) and commercial Monte Carlo (iMC). Seven left-sided breast patients submitted to breast-conserving surgery were enrolled in the study. For each patient, four RT techniques – f-IMRT, IMRT using 2-fields and 5-fields (IMRT2 and IMRT5, respectively) and DCART – were applied. The dose distributions in the planned target volume (PTV) and the dose to the organs at risk (OAR) were compared analyzing dose–volume histograms; further statistical analysis was performed using IBM SPSS v20 software. For PBC, all techniques provided adequate coverage of the PTV. However, statistically significant dose differences were observed between the techniques, in the PTV, OAR and also in the pattern of dose distribution spreading into normal tissues. IMRT5 and DCART spread low doses into greater volumes of normal tissue, right breast, right lung and heart than tangential techniques. However, IMRT5 plans improved distributions for the PTV, exhibiting better conformity and homogeneity in target and reduced high dose percentages in ipsilateral OAR. DCART did not present advantages over any of the techniques investigated. Differences were also found comparing the calculation algorithms: PBC estimated higher doses for the PTV, ipsilateral lung and heart than the iMC algorithm predicted.

Machine Learning Algorithms at Myocardial Perfusion Imaging - a Preliminary Study

Introduction: A major focus of data mining process - especially machine learning researches - is to automatically learn to recognize complex patterns and help to take the adequate decisions strictly based on the acquired data. Since imaging techniques like MPI – Myocardial Perfusion Imaging on Nuclear Cardiology, can implicate a huge part of the daily workflow and generate gigabytes of data, there could be advantages on Computerized Analysis of data over Human Analysis: shorter time, homogeneity and consistency, automatic recording of analysis results, relatively inexpensive, etc.Objectives: The aim of this study relates with the evaluation of the efficacy of this methodology on the evaluation of MPI Stress studies and the process of decision taking concerning the continuation – or not – of the evaluation of each patient. It has been pursued has an objective to automatically classify a patient test in one of three groups: “Positive”, “Negative” and “Indeterminate”. “Positive” would directly follow to the Rest test part of the exam, the “Negative” would be directly exempted from continuation and only the “Indeterminate” group would deserve the clinician analysis, so allowing economy of clinician’s effort, increasing workflow fluidity at the technologist’s level and probably sparing time to patients. Methods: WEKA v3.6.2 open source software was used to make a comparative analysis of three WEKA algorithms (“OneR”, “J48” and “Naïve Bayes”) - on a retrospective study using the comparison with correspondent clinical results as reference, signed by nuclear cardiologist experts - on “SPECT Heart Dataset”, available on University of California – Irvine, at the Machine Learning Repository. For evaluation purposes, criteria as “Precision”, “Incorrectly Classified Instances” and “Receiver Operating Characteristics (ROC) Areas” were considered. Results: The interpretation of the data suggests that the Naïve Bayes algorithm has the best performance among the three previously selected algorithms. Conclusions: It is believed - and apparently supported by the findings - that machine learning algorithms could significantly assist, at an intermediary level, on the analysis of scintigraphic data obtained on MPI, namely after Stress acquisition, so eventually increasing efficiency of the entire system and potentially easing both roles of Technologists and Nuclear Cardiologists. In the actual continuation of this study, it is planned to use more patient information and significantly increase the population under study, in order to allow improving system accuracy.

Modelo de data mining para detecção de tumores em exames de rastreio

Dissertação para obtenção do grau de Mestre em Engenharia Informática