Recently, several public datasets for automatic speech recognition (ASR) in Brazilian Portuguese (BP) have been released, improving ASR systems performance. However, these datasets lack diversity in terms of age groups, regional accents, and education levels. In this paper, we present a new publicly available dataset consisting of 289 life story interviews (365 hours), featuring a broad range of speakers varying in age, education, and regional accents. First, we demonstrated the presence of bias in current BP ASR models concerning education levels and age groups. Second, we showed that our dataset helps mitigate these biases. Additionally, an ASR model trained on our dataset performed better during evaluation on a diverse test set. Finally, the ASR model trained with our dataset was extrinsically evaluated through a topic modeling task that utilized the automatically transcribed output.

The detection of audio deepfakes (ADD) has become increasingly important due to the rapid evolution of generative speech models. However, progress in this field remains uneven across languages, particularly for low-resource languages like Portuguese, which lack high-quality datasets. In this paper, we introduce BRSpeech-DF, the first publicly available ADD dataset for Portuguese, encompassing both Brazilian and European variants. The dataset contains over 458,000 utterances, including a smaller portion of real speech from 62 speakers and a large collection of synthetic samples generated using multiple zero-shot text-to-speech (TTS) models, each conditioned on the original speaker’s voice. By providing this resource, our objective is to support the development of robust, multilingual detection systems, thereby advancing equity in speech forensics and security research. BRSpeech-DF addresses a significant gap in annotated data for underrepresented languages, facilitating more inclusive and generalizable advancements in synthetic speech detection.

This work presents FreeSVC, a promising multilingual singing voice conversion approach that leverages an enhanced VITS model with Speaker-invariant Clustering (SPIN) for better content representation and the State-of-the-Art (SOTA) speaker encoder ECAPA2. FreeSVC incorporates trainable language embeddings to handle multiple languages and employs an advanced speaker encoder to disentangle speaker characteristics from linguistic content. Designed for zero-shot learning, FreeSVC enables cross-lingual singing voice conversion without extensive language-specific training. We demonstrate that a multilingual content extractor is crucial for optimal cross-language conversion. Our source code and models are publicly available1.

In this paper, we propose SC-GlowTTS: an efficient zero-shot multi-speaker text-to-speech model that improves similarity for speakers unseen during training. We propose a speaker-conditional architecture that explores a flow-based decoder that works in a zero-shot scenario. As text encoders, we explore a dilated residual convolutional-based encoder, gated convolutional-based encoder, and transformer-based encoder. Additionally, we have shown that adjusting a GAN-based vocoder for the spectrograms predicted by the TTS model on the training dataset can significantly improve the similarity and speech quality for new speakers. Our model converges using only 11 speakers, reaching state-of-the-art results for similarity with new speakers, as well as high speech quality.

Node-Depth Based Encoding is a representation for Evolutionary Algorithms applied to problems modelled by trees, storing nodes and their respective depths in an appropriately ordered list. This representation was highlighted by the results obtained, whose mutation and recombination operators have low time complexity in relation to other representations for the same problems. This work proposes new search operators. A high locality mutation operator, having as its main differential the possibility of including any edge present in a graph and a recombination operator with the ability to generate solutions with as much heritability as possible considering the edges set in the solutions, making it possible to recombine any solutions in the population with the aim of improving search convergence. All proposed operators also allow for dealing with problems modelled by forests with many trees and complete or non-complete graphs, always generating feasible solutions. This study performed bias, locality and heritability investigations for proposed operators showing that they have adequate characteristics for dealing with high dimensionality problems. In order to evaluate the performance of proposed operators, the results of preliminary tests were obtained applying to a benchmark problem in the literature. The use of proposed operators resulted in faster convergence.

Recommendation systems operate in a highly stochastic and non-stationary environment. As the amount of user-specific information varies, the users’ interests themselves also change. This combination creates a dynamic setting where a single solution will rarely be optimal unless it can keep up with these transformations. One system may perform better than others depending on the situation at hand, thus making the choice of which system to deploy, even more difficult. We address these problems by using the Hierarchical Reinforcement Learning framework. Our proposed meta-bandit acts as a policy over options, where each option maps to a pre-trained, independent recommender system. This meta-bandit learns online and selects a recommender accordingly to the context, adjusting to the situation. We conducted experiments on real data and found that our approach manages to address the dynamics within the user’s changing interests. We also show that it outperforms any of the recommenders separately, as well as an ensemble of them.

Ideally, in FSM-based testing, a test suite should have as few input symbols and sequences as possible, speeding up its execution, and being able to reveal as many faulty implementations as possible in a given fault domain. Unfortunately, in general, it is an NP-hard problem. Several test generation methods have been proposed to generate a test suite with full fault detection capability. However, they usually generate large test suites with many sequences. In this paper, we propose a search-based software testing method aiming to generate test suites that represent a good trade-off between its cost and its fault coverage. A new encoding, called Node-depth-outdegree Encoding (NDOE) and its mutation and initialization operators are proposed. Also, a new fitness function is presented. The proposed fitness is composed of three objectives: minimize the test suite length, minimize the number of test sequences, and maximize the fault coverage, which analysis is based on mutation score. The experimental results show that the proposed method is able to generate lower cost test suites, with high mutation score, than literature test generation methods.

Skin detection based on color can be applied in eHealth systems for preventive healthcare and computer-aided diagnosis. These algorithms could be incorporated in acquisition and preprocessing steps of the applications that assist with skincare, as prevention and detection of melanoma. In this paper we present the results of a study that investigated the reduction of the color spectrum in the HSV system for sample-based skin detection of individuals of different ages and ethnicities. The proposed HSV filter reduced the color spectrum by 97.4648% so as to select candidates for human skin tones. It achieved low sensitivity (54.6333%) and high specificity (92.6390%) in human skin detection in color digital images when compared to the performance of other algorithms proposed in the literature. Different from other filters described in the literature which propose a single interval for human skin in the HSV system, this model presents and discusses 13 intervals in the possible spectrum which present a well-defined variation in terms of tone.

Feature selection is a important tast to reduce dimensionality in large datasets. Datasets from multivariate calibration problems are a good example lf datasets with a large number of features. In literature, there are several types of techniques to reduce the number of features for this problem, among them, evolutionary algorithms such as genetic algorithms (GAs). They have been successfully used with binary encoding to select features in multivariate calibration. However, as far as we know, there is no work in literature which provides an integer encoding GA in such context. Thus, this paper presents an integer-based GA implementation for feature selection in multivariate calibration models. The results demonstrated that our proposal is able to outperform the outcomes of participants from 2014 IDRC regarding model prediction error as well as number of selected features. In this dataset, the samples correspond to oils from petroleum reservoirs around the world and gas mixtures in the gas phase measured in transmittance. The gain of our proposed implementation in relation to the winner was from 20.9% up to 88.8%.

This work proposes a method that allows the entry of text in smartwatches using gestures based on geometric forms. For this it is proposed the development of a prototype capable of inserting a letter with no more than two user interactions. Gesture recognition is performed using the incremental recognition algorithm. A set of gestures with lines and curves were created to be recognized by the incremental recognition algorithm, generated from the reduced equation of the line and the reduced equation of the circumference, respectively. After recognizing the gestures, they are sent to a classifier Naïve Bayes which is responsible for predicting the letter that will be inserted. The Naïve Bayes classifier was trained with a user gesture base that drew all the letters of the alphabet using only the gestures available in the set presented to them. Using the gesture base and the classifier Naïve Bayes a prototype was developed for smartwatches that automatically suggests the most likely letters to be inserted. The prototype was used to perform an experiment, during the experiment the users inserted the five most frequent letters and the five less frequent letters of the English language. The results of the experiment show that the prototype is able to recognize a letter with at most two interactions between the user and the smartwatch. The analysis of the usability and experience test shows that the prototype has generalized potential for use, since it allows the entry of text with up to two interactions and with a 100% hit rate for the most frequent letters and 95,14% For less frequent letters.

Bio-inspired metaheuristics have been used to solve various types of problems. Firefly algorithm (FA) is a nature-inspired metaheuristic based on the flash characteristics of fireflies. Variations of the FA implementation have been applied in the solution of optimization problems, including the problem of variable selection. Some works in the literature have applied several types of strategy in FA to select uncorrelated informative variables. On the other hand, other works have presented the use of compact evolutionary strategy such as compact genetic algorithm (CGA). However, as far as we know, there is still none work that uses such approach in the FA. Therefore, this paper proposes a compact firefly algorithm (CFA) implementation for the variable selection problem in multivariate calibration models to pharmaceutical ingredient determination. In an experiment comparison, our outcomes demonstrate that CFA is able to construct an accurate model with a more reduced number of variables selected. Additionally, CFA is up to 245% better than a standard GA implementation.

This work extends PneumoCAD, a Computer-Aided Diagnosis system for detecting pneumonia in infants using radiographic images, with the aim of improving the system's accuracy and robustness. We implement and compare five con-temporary machine learning classifiers, namely: Naïve Bayes, K-Nearest Neighbor (KNN), Support Vector Machines (SVM), Multi-Layer Perceptron (MLP) and Decision Tree, combined with three dimensionality reduction algorithms: Sequential Forward Selection (SFS), Principal Component Analysis (PCA) and Kernel Principal Component Analysis (KPCA). Current results demonstrate that Naïve Bayes classifier combined with KPCA produces the best overall results.

This paper presents a multi-objective formulation for variable selection in calibration problems. The prediction of protein concentration on wheat is obtained by a linear regression model using variables obtained by a spectrophotometer device. This device measure hundreds of correlated variables related with physicochemical properties and that can be used to estimate the protein concentration. The problem is the selection of a subset informative and uncorrelated variables that help the minimization of prediction error. In this work we propose the use of two objectives in this problem: the prediction error and the number of variables in the model, both related to linear equations system stability. We proposed a multi-objective formulation using two multi-objective algorithms: the NSGA-II and the SPEA-II. Additionally we propose a final decision maker method to choice the final subset of variables from the Pareto front. For the case study is used wheat data obtained by NIR spectrometry where the objective is the determination of a variable subgroup with information about protein concentration. The results of traditional techniques of multivariate calibration as the Successive Projections Algorithm (SPA), Partial Least Square (PLS) and mono-objective genetic algorithm are presents for comparisons. For NIR spectral analysis of protein concentration on wheat, the number of variables selected from 775 spectral variables was reduced for just 10 in the SPEA-II algorithm. The prediction error decreased from 0.2 in the classical methods to 0.09 in proposed approach, a reduction of 45%. The model using variables selected by SPEA-II had better prediction performance than classical algorithms and full-spectrum partial least-squares (PLS).