Elias Teodoro Silva
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Elias Teodoro Silva has a total of 67 co-authors. Here are the top co-authors by number of shared publications:
- Carlos Eduardo Pereira
PPGEE–Graduate Program on Electrical Engineering, Federal University of Rio Grande do Sul, Competence Center on Digital Agriculture - SENAI-RS/EMBRAPII , Porto Alegre, RS 90035190, Brazil
8 shared publications - Flávio Rech Wagner
Institute of Informatics, Federal University of Rio Grande do Sul, Porto Alegre, Brazil
5 shared publications - Pedro Pedrosa Rebouças Filho
Universidade Federal do Ceará
4 shared publications Fausto Sampaio
3 shared publications- Marco Aurelio Wehrmeister
Federal University of Technology – Paraná, Curitiba, Brazil
3 shared publications - Edison Pignaton de Freitas
Informatics Institute, Federal University of Rio Grande Do Sul, Porto Alegre, Brazil
2 shared publications
Latest Publications
- Open Access
- 26 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 170351-170359
Epileptic Seizure Detection Using a Small Memory Footprint Convolutional Neural Network
Automatically detecting epileptic seizures has been explored under the light of machine learning, with promising results in the past few years. Epilepsy affects about 51 million people worldwide, while 30% of them do not have control over their symptoms, which causes unpredictable seizures. This means that approximately 15.3 million people will continue to experience seizure episodes even after undergoing treatment (e.g. drugs, surgery). Many previous works have focused on the development of models for patient-specific tasks. This study will explore the problem of seizure detection under the cross-patient approach. We present a technique for generating and preprocessing data from EEG signals that can be used as inputs to a Convolutional Neural Network. The architecture and performance of the network are presented, with results comparable to the state-of-the-art, achieving around 95% average accuracy. The network presented has five convolutional layers, making it a good candidate to be embedded in a wearable device. The method is explained in detail as well as the training hyperparameters.
- 26 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 170351-170359
Open AccessEpileptic Seizure Detection Using a Small Memory Footprint Convolutional Neural Network
Automatically detecting epileptic seizures has been explored under the light of machine learning, with promising results in the past few years. Epilepsy affects about 51 million people worldwide, while 30% of them do not have control over their symptoms, which causes unpredictable seizures. This means that approximately 15.3 million people will continue to experience seizure episodes even after undergoing treatment (e.g. drugs, surgery). Many previous works have focused on the development of models for patient-specific tasks. This study will explore the problem of seizure detection under the cross-patient approach. We present a technique for generating and preprocessing data from EEG signals that can be used as inputs to a Convolutional Neural Network. The architecture and performance of the network are presented, with results comparable to the state-of-the-art, achieving around 95% average accuracy. The network presented has five convolutional layers, making it a good candidate to be embedded in a wearable device. The method is explained in detail as well as the training hyperparameters.
- Open Access
- 20 July 2025
- Published by Sociedade Brasileira de Computação
- p. 276-285
Computação em Borda para o Monitoramento de Abelhas: Detecção da Ausência da Rainha por Análise de Áudio
A preservação das abelhas é essencial para a biodiversidade, e a presença da abelha-rainha é crucial para a organização da colmeia. Utilizando a computação em borda esse trabalho propõe um método para detectar sua ausência por meio da análise de áudios da colmeia de forma rápida, viável e acessível. As gravações foram processadas, extraindo-se características acústicas relevantes, utilizadas na classificação por Naive Bayes, KNN, MLP e Random Forest. Esses algoritmos alcançaram acurácias que variam de 87,66% até 97,54%. O MLP foi escolhido para implementação em microcontrolador, onde atingiu 88,50% de precisão com um tempo médio de inferência de 109 ms.
- 20 July 2025
- Published by Sociedade Brasileira de Computação
- p. 276-285
Open AccessComputação em Borda para o Monitoramento de Abelhas: Detecção da Ausência da Rainha por Análise de Áudio
A preservação das abelhas é essencial para a biodiversidade, e a presença da abelha-rainha é crucial para a organização da colmeia. Utilizando a computação em borda esse trabalho propõe um método para detectar sua ausência por meio da análise de áudios da colmeia de forma rápida, viável e acessível. As gravações foram processadas, extraindo-se características acústicas relevantes, utilizadas na classificação por Naive Bayes, KNN, MLP e Random Forest. Esses algoritmos alcançaram acurácias que variam de 87,66% até 97,54%. O MLP foi escolhido para implementação em microcontrolador, onde atingiu 88,50% de precisão com um tempo médio de inferência de 109 ms.
- 27 March 2022
- Published by Springer Nature
- p. 1142-1152
Driver Behavior Analysis: Abnormal Driving Detection Using MLP Classifier Applied to Outdoor Camera Images
- 27 March 2022
- Published by Springer Nature
- p. 1142-1152
Driver Behavior Analysis: Abnormal Driving Detection Using MLP Classifier Applied to Outdoor Camera Images
- 1 July 2020
- Published by Elsevier in Microprocessors and Microsystems
- Vol. 76
A method for embedding a computer vision application into a wearable device
- 1 July 2020
- Published by Elsevier in Microprocessors and Microsystems
- Vol. 76
A method for embedding a computer vision application into a wearable device
- Open Access
- 19 November 2019
- Published by Sociedade Brasileira de Computação
- p. 9-16
Arquitetura para o desenvolvimento de soluções embarcadas de internet das coisas aplicadas em redes elétricas inteligentes
As Redes Elétricas Inteligentes (Smart Grids) empregam vários dispositivos para monitorar, analisar e controlar a própria rede. Esses dispositivos são implantados em grande quantidade em usinas de energia, centros de distribuição e nas instalações dos consumidores. Uma Smart Grid requer conectividade, automação e rastreamento para tais dispositivos. Por outro lado, a Internet das Coisas (IoT) tem sido utilizada para prover esses serviços, mas ainda há uma série de desafios a serem superados. Os principais deles se relacionam à alta diversidade de plataformas de hardware e software utilizadas para implementar aqueles computadores embarcados que estão mais próximos da infraestrutura da rede elétrica, chamados dispositivos de borda. Para reduzir o tempo de desenvolvimento dessas aplicações embarcadas, este artigo propõe um middleware IoT, denominado Smart Grids Applications Middleware (SGAM), facilitando a criação de soluções embarcadas para controle, monitoramento e medição em Redes Elétricas Inteligentes. Um caso de uso de monitoramento remoto de ativos foi desenvolvido e é apresentado para demostrar o funcionamento da arquitetura proposta.
- 19 November 2019
- Published by Sociedade Brasileira de Computação
- p. 9-16
Open AccessArquitetura para o desenvolvimento de soluções embarcadas de internet das coisas aplicadas em redes elétricas inteligentes
As Redes Elétricas Inteligentes (Smart Grids) empregam vários dispositivos para monitorar, analisar e controlar a própria rede. Esses dispositivos são implantados em grande quantidade em usinas de energia, centros de distribuição e nas instalações dos consumidores. Uma Smart Grid requer conectividade, automação e rastreamento para tais dispositivos. Por outro lado, a Internet das Coisas (IoT) tem sido utilizada para prover esses serviços, mas ainda há uma série de desafios a serem superados. Os principais deles se relacionam à alta diversidade de plataformas de hardware e software utilizadas para implementar aqueles computadores embarcados que estão mais próximos da infraestrutura da rede elétrica, chamados dispositivos de borda. Para reduzir o tempo de desenvolvimento dessas aplicações embarcadas, este artigo propõe um middleware IoT, denominado Smart Grids Applications Middleware (SGAM), facilitando a criação de soluções embarcadas para controle, monitoramento e medição em Redes Elétricas Inteligentes. Um caso de uso de monitoramento remoto de ativos foi desenvolvido e é apresentado para demostrar o funcionamento da arquitetura proposta.
- 1 November 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 172-177
Evaluating Resources Cost of a Convolutional Neural Network Aiming an Embedded System
The use of Machine Learning algorithms in image classification problems have yielded satisfactory results in recent years. Classification algorithms such as Support Vector Machines (SVMs) combined with robust feature extractors like Histogram of Oriented Gradients (HOG) have been used to achieve accuracy results over 95%. Very recently, with the researches applied in the deep learning fields, Convolutional Neural Networks (CNNs) have shown to work extremely well with data that has high dimensionality like images. This work focuses on evaluating the resources costs of deploying a CNN in an embedded platform to solve the people detection problem. An implementation of the CNN classification algorithm was developed, and tests were carried out both in a PC and in an embedded platform. Furthermore, a study on the amount of memory and time spent by a classic CNN was executed. The results point out that new networks must be designed to fit in low-resources embedded platforms.
- 1 November 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 172-177
Evaluating Resources Cost of a Convolutional Neural Network Aiming an Embedded System
The use of Machine Learning algorithms in image classification problems have yielded satisfactory results in recent years. Classification algorithms such as Support Vector Machines (SVMs) combined with robust feature extractors like Histogram of Oriented Gradients (HOG) have been used to achieve accuracy results over 95%. Very recently, with the researches applied in the deep learning fields, Convolutional Neural Networks (CNNs) have shown to work extremely well with data that has high dimensionality like images. This work focuses on evaluating the resources costs of deploying a CNN in an embedded platform to solve the people detection problem. An implementation of the CNN classification algorithm was developed, and tests were carried out both in a PC and in an embedded platform. Furthermore, a study on the amount of memory and time spent by a classic CNN was executed. The results point out that new networks must be designed to fit in low-resources embedded platforms.
- 1 November 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 46-53
An Embedded Application to Identify Degradation in Energized Polymeric Insulators Using Machine Learning and Wavelet Transform
Insulators are among the main causes of failure in the electric power lines. In this paper is described the process to develop and embed an application to identify degradation in high voltage polymeric insulators using ultrasonic emissions. The proposed approach is a combination of the wavelet transform and two different classifiers, Naive Bayes or Knn. Both strategies are evaluated in a workstation and in an embedded platform, an ARM Cortex M4. Their accuracy, execution time, and memory footprint are compared for the embedded implementation. The results indicate that the selected techniques offer good prediction rate and can be embedded in low-cost microcontrollers.
- 1 November 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 46-53
An Embedded Application to Identify Degradation in Energized Polymeric Insulators Using Machine Learning and Wavelet Transform
Insulators are among the main causes of failure in the electric power lines. In this paper is described the process to develop and embed an application to identify degradation in high voltage polymeric insulators using ultrasonic emissions. The proposed approach is a combination of the wavelet transform and two different classifiers, Naive Bayes or Knn. Both strategies are evaluated in a workstation and in an embedded platform, an ARM Cortex M4. Their accuracy, execution time, and memory footprint are compared for the embedded implementation. The results indicate that the selected techniques offer good prediction rate and can be embedded in low-cost microcontrollers.
- 1 July 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Estimation of global solar irradiance with LDR sensor and artificial neural network embedded in an 8-bit microcontroller
This work deals with the estimation of global solar irradiance in an embedded platform through the use of a Light Dependent Resistor (LDR) sensor and its temperature. A prototype is built and has its values compared to the data obtained through a commercial pyranometer. An Multilayer Perceptron neural network is used to provide a non-linear regression between the voltage and temperature signals of the LDR on the commercial sensor irradiance data in a data analysis software. A regression with determination coefficient of 96.466 % and with mean squared error of 0.04 was obtained. The neural network present in the embedded system has 100% accuracy in relation to the neural network present in the data analysis software used. The minimum response time of the prototype is 13.49 ms and its dissipated power is 27.2 mW, making the approach quite promising. This neural estimation, using simple sensors like the LDR, can help to reduce the cost of renewable energy applications and make it easier to implement.
- 1 July 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Estimation of global solar irradiance with LDR sensor and artificial neural network embedded in an 8-bit microcontroller
This work deals with the estimation of global solar irradiance in an embedded platform through the use of a Light Dependent Resistor (LDR) sensor and its temperature. A prototype is built and has its values compared to the data obtained through a commercial pyranometer. An Multilayer Perceptron neural network is used to provide a non-linear regression between the voltage and temperature signals of the LDR on the commercial sensor irradiance data in a data analysis software. A regression with determination coefficient of 96.466 % and with mean squared error of 0.04 was obtained. The neural network present in the embedded system has 100% accuracy in relation to the neural network present in the data analysis software used. The minimum response time of the prototype is 13.49 ms and its dissipated power is 27.2 mW, making the approach quite promising. This neural estimation, using simple sensors like the LDR, can help to reduce the cost of renewable energy applications and make it easier to implement.
- 1 June 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Toward an Embedded Application to Identify Degradation in Energized Polymeric Insulators using Discrete Wavelet Transform
This paper presents an approach to distinguish among different aging conditions in insulators present on power distribution nets. The ultrasound emitted by insulators, when submitted to the nominal voltage, is captured and processed by Discrete Wavelet Transform. Two classifiers are used, Bayes and KNN, obtaining about 93 % of accuracy. It also analyses the execution time and memory footprint of a DWT implemented in an embedded platform (ARM Cortex-M). In the end, a comparison is made between the computational simulation and embedded accuracies.
- 1 June 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Toward an Embedded Application to Identify Degradation in Energized Polymeric Insulators using Discrete Wavelet Transform
This paper presents an approach to distinguish among different aging conditions in insulators present on power distribution nets. The ultrasound emitted by insulators, when submitted to the nominal voltage, is captured and processed by Discrete Wavelet Transform. Two classifiers are used, Bayes and KNN, obtaining about 93 % of accuracy. It also analyses the execution time and memory footprint of a DWT implemented in an embedded platform (ARM Cortex-M). In the end, a comparison is made between the computational simulation and embedded accuracies.
- 11 May 2018
- Published by Springer Nature
- p. 201-213
An Embedded Classifier for Mobile Robot Localization Using Support Vector Machines and Gray-Level Co-occurrence Matrix
- 11 May 2018
- Published by Springer Nature
- p. 201-213
An Embedded Classifier for Mobile Robot Localization Using Support Vector Machines and Gray-Level Co-occurrence Matrix