Andres Gomez-Bautista
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Andres Gomez-Bautista has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Catalina Alvarado-Rojas
School of Engineering, Pontificia Universidad Javeriana, Bogota 110231, Colombia
3 shared publications - Diego Mendez
Electronics Engineering Department, Pontificia Universidad Javeriana, Bogotá 110231, Colombia
3 shared publications - Julian Colorado
Pontificia Universidad Javeriana, Bogotá, Colombia
3 shared publications - Fredy Cuellar
Electronic Engineering Department, Pontifical Javeriana University, Bogot, Colombia
2 shared publications - John E. Bermeo
Instituto Politécnico Nacional, Centro de Investigación en Ciencia Aplicada y Tecnología Avanzada Unidad Querétaro
2 shared publications - Francisco Calderon
Universidad Tecnológica Nacional, Facultad Regional Mendoza, Centro Regional de Desarrollos Tecnológicos para la Construcción, Sismología e Ingeniería Sísmica (CeReDeTeC), Argentina
1 shared publication
Latest Publications
- Open Access
On the Deployment of Edge AI Models for Surface Electromyography-Based Hand Gesture Recognition
Background: Robotic-based therapy has emerged as a prominent treatment modality for the rehabilitation of hand function impairment resulting from strokes. Aim: In this context, feature engineering becomes particularly important to estimate the intention of upper limb movements by utilizing machine learning models, especially when a hardware embedded-on-board implementation is expected, due to the strong computational, energy, and latency constraints. Methods: The present study details the implementation of four cutting-edge feature engineering techniques (random forest, minimum redundancy maximum relevance (MRMR), Davies–Bouldin index, and t-tests) in the context of machine learning algorithms (neuronal networks and bagged forests) deployed within a resource-constrained autonomous embedded system. Results: The findings of this study demonstrate that by assigning relative importance to features and removing redundant or superfluous information, it is possible to enhance the system’s execution by up to 31% while preserving the model’s performance at a comparable level. Conclusions: This work proves the usefulness of TinyML as an approach to properly integrate AI into constrained edge embedded systems to support complex strategies such as the proposed hand gesture recognition for the smart rehabilitation of post-stroke patients.
Open AccessOn the Deployment of Edge AI Models for Surface Electromyography-Based Hand Gesture Recognition
Background: Robotic-based therapy has emerged as a prominent treatment modality for the rehabilitation of hand function impairment resulting from strokes. Aim: In this context, feature engineering becomes particularly important to estimate the intention of upper limb movements by utilizing machine learning models, especially when a hardware embedded-on-board implementation is expected, due to the strong computational, energy, and latency constraints. Methods: The present study details the implementation of four cutting-edge feature engineering techniques (random forest, minimum redundancy maximum relevance (MRMR), Davies–Bouldin index, and t-tests) in the context of machine learning algorithms (neuronal networks and bagged forests) deployed within a resource-constrained autonomous embedded system. Results: The findings of this study demonstrate that by assigning relative importance to features and removing redundant or superfluous information, it is possible to enhance the system’s execution by up to 31% while preserving the model’s performance at a comparable level. Conclusions: This work proves the usefulness of TinyML as an approach to properly integrate AI into constrained edge embedded systems to support complex strategies such as the proposed hand gesture recognition for the smart rehabilitation of post-stroke patients.
- Open Access
Scalable O(log2n) Dynamics Control for Soft Exoskeletons
Robotic exoskeletons are being actively applied to support the activities of daily living (ADL) for patients with hand motion impairments. In terms of actuation, soft materials and sensors have opened new alternatives to conventional rigid body structures. In this arena, biomimetic soft systems play an important role in modeling and controlling human hand kinematics without the restrictions of rigid mechanical joints while having an entirely deformable body with limitless points of actuation. In this paper, we address the computational limitations of modeling large-scale articulated systems for soft robotic exoskeletons by integrating a parallel algorithm to compute the exoskeleton’s dynamics equations of motion (EoM), achieving a computation with complexity for the highly articulated n degrees of freedom (DoF) running on p processing cores. The proposed parallel algorithm achieves an exponential speedup for DoF while achieving a degree of parallelism for , which demonstrates the required scalability for controlling highly articulated soft exoskeletons in real time. However, scalability will be bounded by the fraction.Open AccessScalable O(log2n) Dynamics Control for Soft Exoskeletons
Robotic exoskeletons are being actively applied to support the activities of daily living (ADL) for patients with hand motion impairments. In terms of actuation, soft materials and sensors have opened new alternatives to conventional rigid body structures. In this arena, biomimetic soft systems play an important role in modeling and controlling human hand kinematics without the restrictions of rigid mechanical joints while having an entirely deformable body with limitless points of actuation. In this paper, we address the computational limitations of modeling large-scale articulated systems for soft robotic exoskeletons by integrating a parallel algorithm to compute the exoskeleton’s dynamics equations of motion (EoM), achieving a computation with complexity for the highly articulated n degrees of freedom (DoF) running on p processing cores. The proposed parallel algorithm achieves an exponential speedup for DoF while achieving a degree of parallelism for , which demonstrates the required scalability for controlling highly articulated soft exoskeletons in real time. However, scalability will be bounded by the fraction. - Open Access
- 29 September 2024
- Published by MDPI AG in Preprints.org
Scalable O(log2n) Dynamics Control for Soft Exoskeletons
Robotic exoskeletons are being actively applied to support the activities of daily living (ADL) for patients with hand motion impairments. In terms of actuation, soft materials and sensors have opened new alternatives upon conventional rigid body structures. In this arena, biomimetic soft systems are playing an important role for modeling and controlling the human hand kinematics without the restrictions of rigid mechanical joints, while having an entire deformable body with limitless points of actuation. In this paper, we address the computational limitations for modeling large-scale articulated systems for soft robotic exoskeletons, by integrating a parallel architecture to compute the exoskeleton’s dynamics equations of motion (EoM), obtaining a computation with O(log2n) complexity for the highly-articulated n degrees of freedom.
- 29 September 2024
- Published by MDPI AG in Preprints.org
Open AccessScalable O(log2n) Dynamics Control for Soft Exoskeletons
Robotic exoskeletons are being actively applied to support the activities of daily living (ADL) for patients with hand motion impairments. In terms of actuation, soft materials and sensors have opened new alternatives upon conventional rigid body structures. In this arena, biomimetic soft systems are playing an important role for modeling and controlling the human hand kinematics without the restrictions of rigid mechanical joints, while having an entire deformable body with limitless points of actuation. In this paper, we address the computational limitations for modeling large-scale articulated systems for soft robotic exoskeletons, by integrating a parallel architecture to compute the exoskeleton’s dynamics equations of motion (EoM), obtaining a computation with O(log2n) complexity for the highly-articulated n degrees of freedom.
- Open Access
- 8 February 2024
- Published by Universidad de Antioquia in Revista Facultad de Ingeniería Universidad de Antioquia
Enhancing Facial Recognition in Surveillance Systems through Embedded Super-resolution
Este documento detalla la implementación de una red neuronal convolucional de aplicación a nivel sub-pixel diseñada para mejorar la resolución de imágenes faciales. El modelo utiliza una serie de filtros para aumentar progresivamente el número de píxeles, estimando la información necesaria para los nuevos píxeles tanto de la imagen original como del entrenamiento derivado de 17,500 imágenes sintéticas producidas por redes neuronales adversarias. Dentro del contexto de la vigilancia y aplicaciones relacionadas, la red neuronal convolucional entrenada muestra características beneficiosas. Por ejemplo, se puede implementar dentro de un dispositivo para lograr imágenes de mayor resolución de las que la cámara física puede producir. Esta investigación subraya la viabilidad de dicho dispositivo a través de la implementación y evaluación de la red en el sistema embebido NVIDIA Jetson TX2. Los hallazgos demuestran la practicidad del modelo para aplicaciones de vigilancia en tiempo real y su capacidad para producir imágenes de calidad superior en comparación con varios métodos de interpolación, según lo determinado por un proceso de prueba exhaustivo que mide varios atributos de las imágenes generadas.
- 8 February 2024
- Published by Universidad de Antioquia in Revista Facultad de Ingeniería Universidad de Antioquia
Open AccessEnhancing Facial Recognition in Surveillance Systems through Embedded Super-resolution
Este documento detalla la implementación de una red neuronal convolucional de aplicación a nivel sub-pixel diseñada para mejorar la resolución de imágenes faciales. El modelo utiliza una serie de filtros para aumentar progresivamente el número de píxeles, estimando la información necesaria para los nuevos píxeles tanto de la imagen original como del entrenamiento derivado de 17,500 imágenes sintéticas producidas por redes neuronales adversarias. Dentro del contexto de la vigilancia y aplicaciones relacionadas, la red neuronal convolucional entrenada muestra características beneficiosas. Por ejemplo, se puede implementar dentro de un dispositivo para lograr imágenes de mayor resolución de las que la cámara física puede producir. Esta investigación subraya la viabilidad de dicho dispositivo a través de la implementación y evaluación de la red en el sistema embebido NVIDIA Jetson TX2. Los hallazgos demuestran la practicidad del modelo para aplicaciones de vigilancia en tiempo real y su capacidad para producir imágenes de calidad superior en comparación con varios métodos de interpolación, según lo determinado por un proceso de prueba exhaustivo que mide varios atributos de las imágenes generadas.