Showing papers by "Diego B. Haddad published in 2021"

Flattening the curves: on-off lock-down strategies for COVID-19 with an application to Brazil

Abstract: The current COVID-19 pandemic is affecting different countries in different ways. The assortment of reporting techniques alongside other issues, such as underreporting and budgetary constraints, makes predicting the spread and lethality of the virus a challenging task. This work attempts to gain a better understanding of how COVID-19 will affect one of the least studied countries, namely Brazil. Currently, several Brazilian states are in a state of lock-down. However, there is political pressure for this type of measures to be lifted. This work considers the impact that such a termination would have on how the virus evolves locally. This was done by extending the SEIR model with an on / off strategy. Given the simplicity of SEIR we also attempted to gain more insight by developing a neural regressor. We chose to employ features that current clinical studies have pinpointed has having a connection to the lethality of COVID-19. We discuss how this data can be processed in order to obtain a robust assessment.

Diagnostic and severity analysis of combined failures composed by imbalance and misalignment in rotating machines

Abstract: Failure detection from mechanical vibration analysis is crucial in industry machinery, with early discovery allowing for preventive action to be performed. This paper introduces a prototype of an IoT system capable of (i) identifying combined failures of a rotating machine and (ii) predicting failures, in a non-invasive manner. An embedded solution is devised, which is able to classify four types of operating conditions, namely (i) normal, (ii) imbalanced, (iii) imbalanced associated with horizontal misalignment, and (iv) imbalanced associated with vertical misalignment. The goal of the paper is to propose an automatic method of diagnosis and measurement of combined failures in rotating machines. The employed methodology combines a simulation bench and measuring the severity in a controlled environment. Three distinct machine learning techniques were compared for classification purposes: support vector machines, k-nearest neighbors, and random forests. The results obtained reveal the possibility of differentiating between the types of combined faults; an accuracy of 81.41% using a random forest classifier was achieved. A supervisory system was developed which is responsible for monitoring machines and sending wireless alert messages. The latter are sent to a control application, allowing for user interaction through mobile devices. Results reveal the possibility of differentiating between the types of combined faults, and also motor failure severity profile for different scenarios. Through the construction of severity profiles, when faults occurred, high vibration values were registered at elevated speeds. The proposed methodology can be used in any rotating machine that complies with the conditions imposed by ISO 10816.

A Family of Adaptive Volterra Filters Based on Maximum Correntropy Criterion for Improved Active Control of Impulsive Noise

Abstract: Linear active noise control (ANC) systems have been used in the past to effectively suppress Gaussian noise. A practical ANC system must consider nonlinearities in the secondary path with a non-minimum phase. For an ANC system to effectively operate in a modern-day acoustic environment with multiple electrical/electronic systems operating in the vicinity, the reference noise source is taken as a non-Gaussian stochastic process. Linear systems have shown unacceptable performance in countering the disturbances that are impulsive in nature. Considering these issues, we propose a family of adaptive algorithms for ANC systems that employ a second-order Volterra filter for accurate modeling of the impulsive disturbances. We utilize the maximum correntropy criterion as the cost function to improve the adaptive filtering process, which generates the most appropriate output signal in the ANC system’s successive iterative stages. The proposed algorithms feature dynamic learning-rate parameters, which improve the tracking performance of the algorithms. Also, a careful selection of the Volterra filter’s kernel size in the proposed algorithms ensures a balance between system stability and convergence rate. These parameters are made automatically adjustable, based on the residual error and the reference input signals, to optimize performance in non-Gaussian environments. A comparison of the proposed algorithms with its counterparts is presented through computer simulations in terms of average noise reduction for different levels of impulsive noise. The proposed algorithms are also compared with each other to find the best performing algorithm in a non-Gaussian environment. Achieved results exhibit the effectiveness of the proposed algorithms in their ability to attenuate impulsive noise in comparison with the existing adaptive algorithms for ANC systems. Later, the proposed algorithm has been implemented for speech enhancement, where the noise from the speech sample is identified and removed using an adaptive filter algorithm. Simulations reveal that the devised algorithm can effectively denoise speech signals. Further, in the last part of the paper, the VF-MCCRMC is modified in accordance with the energy of the error signal, and the simulation results demonstrated the improvement in stability and error performance.

Transient Analysis of the Set-Membership LMS Algorithm

Abstract: Adaptive filtering algorithms, which adopt the set-membership strategy, are able to attain good steady-state performance with low computational burden. In general, such advantages are obtained by defining a bounded-error. This specification translates into a time-variant step size, chosen in each iteration according to a nonlinear function of the instantaneous error. Unfortunately, this type of nonlinear behaviour hampers the stochastic modelling of these algorithms. This work devises a novel transient analysis of the Set-Membership Least Mean Squares algorithm. Additionally, a new interpretation is advanced about the implicit optimization problem solved by the algorithm. This explanation is important since it can contribute to the design of new adaptive algorithms. The proposed theoretical analysis provides predictions for: (i) steady-state performance; (ii) transient performance; and (iii) evolution of the update probability. It is noteworthy that the latter influences the computational complexity of the algorithm. Furthermore, we perform a novel comprehensive transient analysis of a set-membership algorithm. In addition, both time-variant transfer functions and deficient-length configurations are addressed. The resulting theoretical estimates are confirmed by simulations.

Análise Comparativa de Métodos para Esteganografia Digital em Imagens

Abstract: Sensitive information being shared on the internet is growing. Because of this, it is increasingly necessary to take security measures whilst this information travels in the network. Digital steganography allows one to send sensitive information in a hidden manner. Although there is a plethora of techniques for such a goal, finding an appropriate one is not always simple. This paper implements and compares spatial-domain digital steganography techniques in both RGB and grayscale images. A frequency-domain heuristic for reducing the visual impact of digital steganography in grayscale images is presented. As another result of this work, a dataset is also available in the Kaggle platform with 18 GB of images, containing secret messages using the techniques under study. In addition, a Python language library was also made available in the PyPI repository, allowing for both concealment and revelation of messages using the presented digital steganography methods.

A Rhythmic Activation Mechanism for Soft Multi-legged Robots

Abstract: Compared to standard solutions, soft robotics presents enhanced adaptability to unpredictable and unstructured environments, encompassing advances in fabrication, modeling, and control. The absence of a general theory for the latter is one of the biggest challenges in the field, which constrains these robots’ employment in real-world applications. This research proposes the application of Scheduling by Multiple Edge Reversal (SMER) in the activation of soft legs to be applied in multi-legged robots. A soft device was developed to be tested as a robot’s leg to evaluate the proposed application. A logic controller for this device was designed using the SMER technique. Image processing techniques were used to assess the functionality of the proposed strategy, which demands limited resources. The vision tracking system is composed of a set of infrared-reflective patches, an infrared illuminator, and a pair of cameras with no infrared filters. Results revealed that it is possible to use SMER techniques to activate soft robotics systems and that the methods employed to develop and test the device were appropriate.

On the Skewness of the LMS Adaptive Weights

Abstract: The adjustable weights of adaptive filtering algorithms are usually assumed to obey a Gaussian distribution. This is somewhat natural under maximal-entropy considerations, since most analyses in the open literature only take into account first-and second-order statistics. This work investigates the third-order statistical feature known as skewness of the least mean square parameters distribution. Two theoretical analyses for skewness estimation are proposed: i ) one that employs the independence assumption, which states that the excitation data is statistically independent from the adaptive weights; ii ) one derived from the exact expectation analysis, a method that is able to predict the learning capabilities of the least mean square algorithm even when the step size is not infinitesimally small. This brief shows that the skewness of the adaptive weights distribution may present a large deviation from the common Gaussian assumption, especially in the first phase of the learning. Furthermore, it is also demonstrated that the skewness may grow without limit even when adaptive weights present convergence in both average and mean square behaviors.

On the Analysis of the Incremental $$\ell _0$$ ℓ 0 -LMS Algorithm for Distributed Systems

Abstract: Adaptive filtering algorithms implement an estimation of a set of parameters. Frequently, the system to be identified is sparse, in the sense that most of its energy is concentrated among a few coefficients. Adaptive algorithms, such as the $$\ell _0$$ -LMS, can incorporate this property in order to increase the convergence rate. In this work, a stochastic model is proposed to predict characteristics of transient, steady state and tracking of the $$\ell _0$$ -LMS algorithm, implemented in a distributed way (i.e., with the incremental strategy). Such a diffuse strategy is adequate in situations where the network energy is severely limited. The advanced analysis does not require neither white nor Gaussian input signals in order to predict the learning capabilities of the $$\ell _0$$ -LMS algorithm.

Blind and Semi-blind Anechoic Mixing System Identification Using Multichannel Matching Pursuit

Abstract: Sparse component analysis techniques have been successfully applied to the separation of speech sources. This paper presents an efficient algorithm based on the matching pursuit approach to deal with multichannel records. The proposed algorithm explicitly employs spatial constraints among different channels to express mixed signals as linear combinations of delayed components selected from an overcomplete dictionary. We present a new procedure for estimating the mixing system parameters (attenuations and delays), which can be applied to more than two mixtures and is not restricted to non-negative attenuation coefficients. The proposed mixing system estimation method can accommodate delays of greater magnitude than traditional approaches. In addition, learned dictionaries that improve the identification step can be used when excerpts from sources (exogenous to mixtures) are available. The simulation results show that semi-blind dictionaries perform better than those used in blind configurations.

Transient Analysis of the Block Least Mean Squares Algorithm

Abstract: It is known that adaptive filtering algorithms may tackle relevant communication tasks. In order to reduce the adaptation rate, the least mean squares algorithm and its normalized version may be implemented in a block manner, so that the filter coefficients are adjusted once per each output block. This letter advances a stochastic model that is able to predict the learning capabilities of time-domain block extensions of these algorithms, and demonstrates that their behaviour is not governed by trivial generalizations of the rules presented by standard implementations. The devised model decouples the radial and angular distribution of input data for the sake of emphasizing the factors that drive the algorithms learning behaviour. Both algorithms are demonstrated to solve a local and deterministic optimization problem. This novel point of view is employed to derive new versions of these algorithms that are able to enhance asymptotic performance by the usage of coefficient reusing techniques. Theoretical results reveal good adherence to simulated learning curves and the proposed algorithms outperform the standard ones in steady-state.

Constrained Least Mean Square Algorithm with Coefficient Reusing

Abstract: The constrained least mean square (CLMS) algorithm is one of the most popular online linear-equality-constrained beamforming algorithms. This paper demonstrates for the first time that it solves a deterministic minimum-disturbance optimization problem in an exact manner. Such a framework is employed to insert the coefficient reusing technique into the algorithm, engendering a new low-complexity constrained adaptive filter, designated as RC-CLMS, that trades convergence rate for asymptotic performance. A stochastic model that predicts the average evolution of adaptive weights is derived. Through simulations, the advanced reusing coefficient extension of the constrained least mean-square algorithm enhanced the asymptotic signal-to-interference-plus-noise ratio and decreased the steady-state mean output energy. Furthermore, the resulting beam pattern is analyzed with an antenna analysis tool, confirming the efficacy of the advanced algorithm in a realistic setting, when the electromagnetic coupling between the antennas is taken into account.