Víctor Hugo Benítez Baltazar
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Víctor Hugo Benítez Baltazar has a total of 54 co-authors. Here are the top co-authors by number of shared publications:
- Jesus Pacheco
Department of Industrial Engineering, Universidad de Sonora, Hermosillo 83000, Mexico
11 shared publications - Luis Carlos Felix-Herran
School of Engineering and Sciences, Tecnológico de Monterrey, Av. Eugenio Garza Sada 2501 Sur, Monterrey 64700, NL, Mexico
9 shared publications - Agustín Brau Ávila
University of Sonora
5 shared publications - Jorge de Jesus Lozoya Santos
Instituto Tecnológico de Monterrey
5 shared publications - Enrique A. Enríquez-Velásquez
School of Engineering, The University of Edinburgh, Sanderson Building, Robert Stevenson Road, The King’s Buildings, Edinburgh EH9 3FB, UK
3 shared publications - Guillermo Pérez
Departamento de Ingeniería Industrial, Campus Hermosillo, Universidad de Sonora, Hermosillo 83000, Mexico
3 shared publications
Latest Publications
- Open Access
- 12 November 2026
- Published by MDPI AG in Electronics
- Vol. 15 (22) , 5270
Timing Feasibility of Multimodal Software-Defined Haptic Rendering with a Browser-Based Digital Twin on a Cortex-M4 Microcontroller
Haptic interfaces convey force between an operator and a physical or virtual environment, yet passive devices encode a single force behavior in dedicated mechanical hardware and cannot emulate qualitatively different phenomena without physical modification. This study presents a software-defined haptic platform that renders four models on a single low-cost 32-bit microcontroller without hardware reconfiguration. A deterministic 1 kHz interrupt-driven control loop drives a motorized linear slider through an H-bridge and renders virtual detents, square-root position-dependent resistance, a one-dimensional reaction–diffusion field solved by explicit finite differences, and a symmetric square-root spring. A browser-based three-dimensional visualization, synchronized with the embedded state over a one-way serial link, forms a shadow Digital Twin, and a two-level limit-switch layer provides end-of-travel protection. The loop period, the worst-case execution time of the field solver, and the telemetry interval were measured on the deployed firmware. The control loop held its 1 ms period with microsecond-level jitter while streaming telemetry to the Digital Twin at approximately 5 Hz, and the field solver consumed less than 8% of the interrupt budget. These results show that reconfigurable multimodal haptic rendering can be scheduled deterministically on resource-constrained embedded hardware, whereas the physical fidelity of the rendered behaviors remains to be validated.
- 12 November 2026
- Published by MDPI AG in Electronics
- Vol. 15 (22) , 5270
Open AccessTiming Feasibility of Multimodal Software-Defined Haptic Rendering with a Browser-Based Digital Twin on a Cortex-M4 Microcontroller
Haptic interfaces convey force between an operator and a physical or virtual environment, yet passive devices encode a single force behavior in dedicated mechanical hardware and cannot emulate qualitatively different phenomena without physical modification. This study presents a software-defined haptic platform that renders four models on a single low-cost 32-bit microcontroller without hardware reconfiguration. A deterministic 1 kHz interrupt-driven control loop drives a motorized linear slider through an H-bridge and renders virtual detents, square-root position-dependent resistance, a one-dimensional reaction–diffusion field solved by explicit finite differences, and a symmetric square-root spring. A browser-based three-dimensional visualization, synchronized with the embedded state over a one-way serial link, forms a shadow Digital Twin, and a two-level limit-switch layer provides end-of-travel protection. The loop period, the worst-case execution time of the field solver, and the telemetry interval were measured on the deployed firmware. The control loop held its 1 ms period with microsecond-level jitter while streaming telemetry to the Digital Twin at approximately 5 Hz, and the field solver consumed less than 8% of the interrupt budget. These results show that reconfigurable multimodal haptic rendering can be scheduled deterministically on resource-constrained embedded hardware, whereas the physical fidelity of the rendered behaviors remains to be validated.
- Open Access
- 14 October 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (20) , 10161
Vision-Based Measurement System for Automated Quality Inspection of CNC Turned Components
Manual inspection in the high-precision manufacturing industry often struggles to meet the demands of high-volume production and strict dimensional tolerances. This paper presents a vision-based measurement system for the automated inspection of CNC-turned parts. The methodology integrates a custom-built, isolated backlit inspection station and sub-pixel edge detection algorithms developed in Python/OpenCV. Optical calibration was configured through a full factorial design of experiments, yielding a ground sample distance of 0.0323 mm/px and a mean reprojection error of 0.093 px. Validation against a coordinate measuring machine on a 108-part batch showed a root mean square error below 0.0260 mm for all dimensions. Error sources were quantified through stability tests, a gauge R&R study, and an uncertainty budget. Overall classification accuracy was 99.1%, 95.4%, and 92.6% at tolerances of ±0.0762, ±0.0508, and ±0.0254 mm, respectively, with sensitivity of at least 97.4% and positive predictive value between 76.2% and 92.6%. The system inspects 100% of the parts, evaluating five dimensions in 4.72 s per piece, over 81% faster than manual inspection, with full digital traceability.- 14 October 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (20) , 10161
Open AccessVision-Based Measurement System for Automated Quality Inspection of CNC Turned Components
Manual inspection in the high-precision manufacturing industry often struggles to meet the demands of high-volume production and strict dimensional tolerances. This paper presents a vision-based measurement system for the automated inspection of CNC-turned parts. The methodology integrates a custom-built, isolated backlit inspection station and sub-pixel edge detection algorithms developed in Python/OpenCV. Optical calibration was configured through a full factorial design of experiments, yielding a ground sample distance of 0.0323 mm/px and a mean reprojection error of 0.093 px. Validation against a coordinate measuring machine on a 108-part batch showed a root mean square error below 0.0260 mm for all dimensions. Error sources were quantified through stability tests, a gauge R&R study, and an uncertainty budget. Overall classification accuracy was 99.1%, 95.4%, and 92.6% at tolerances of ±0.0762, ±0.0508, and ±0.0254 mm, respectively, with sensitivity of at least 97.4% and positive predictive value between 76.2% and 92.6%. The system inspects 100% of the parts, evaluating five dimensions in 4.72 s per piece, over 81% faster than manual inspection, with full digital traceability. - Open Access
- 21 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open-source ESP32-CAM automated image capture chamber for chickpea (Cicer arietinum) quality classification
This article details the design, construction, and validation of an open-source, automated image capture platform based on an ESP32-CAM module, tailored specifically for chickpea (Cicer arietinum) quality classification. The low-cost hardware features a manually adjustable Z-axis mechanism to set a precise, reproducible focal distance, complemented by a controlled dual-illumination system, integrating a custom LED mesh and a bottom backlight to standardize image acquisition and minimize shadowing. Using this platform, a comprehensive dataset comprising 283 images was collected and distributed across four quality categories: intact (healthy), mechanically damaged, humidity-affected, and foreign materials (stones). Spatial illumination uniformity was quantitatively evaluated, yielding acceptable coefficients of variation (CV < 20%). Color and texture statistical descriptors (means, standard deviations, and RMS across RGB channels) were extracted to train two Classification and Regression Tree (CART) pipelines, each independently enhanced by a different large language model (Gemini and ChatGPT) as part of a controlled comparison of LLM-assisted code development for embedded vision systems. Rigorous evaluation protocols were implemented, including stratified 5-fold cross-validation (achieving a stable mean accuracy of 95.1% ± 2.2%) and a completely blind test set comprising unseen captures. The system achieved an overall classification accuracy of up to 0.99 under optimized configurations. Complete mechanical CAD files, electronic schematics, Bill of Materials (BOM), and Python-based classification scripts are publicly hosted in open repositories under CERN-OHL-W, MIT, and CC-BY-4.0 licenses. This reproducible platform provides a cost-effective, accessible alternative to commercial optical sorters for academic research and small-scale agro-industrial applications.
- 21 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessOpen-source ESP32-CAM automated image capture chamber for chickpea (Cicer arietinum) quality classification
This article details the design, construction, and validation of an open-source, automated image capture platform based on an ESP32-CAM module, tailored specifically for chickpea (Cicer arietinum) quality classification. The low-cost hardware features a manually adjustable Z-axis mechanism to set a precise, reproducible focal distance, complemented by a controlled dual-illumination system, integrating a custom LED mesh and a bottom backlight to standardize image acquisition and minimize shadowing. Using this platform, a comprehensive dataset comprising 283 images was collected and distributed across four quality categories: intact (healthy), mechanically damaged, humidity-affected, and foreign materials (stones). Spatial illumination uniformity was quantitatively evaluated, yielding acceptable coefficients of variation (CV < 20%). Color and texture statistical descriptors (means, standard deviations, and RMS across RGB channels) were extracted to train two Classification and Regression Tree (CART) pipelines, each independently enhanced by a different large language model (Gemini and ChatGPT) as part of a controlled comparison of LLM-assisted code development for embedded vision systems. Rigorous evaluation protocols were implemented, including stratified 5-fold cross-validation (achieving a stable mean accuracy of 95.1% ± 2.2%) and a completely blind test set comprising unseen captures. The system achieved an overall classification accuracy of up to 0.99 under optimized configurations. Complete mechanical CAD files, electronic schematics, Bill of Materials (BOM), and Python-based classification scripts are publicly hosted in open repositories under CERN-OHL-W, MIT, and CC-BY-4.0 licenses. This reproducible platform provides a cost-effective, accessible alternative to commercial optical sorters for academic research and small-scale agro-industrial applications.
- 6 October 2025
- Published by Astra Ediciones
- p. 213-223
Humanizing Fault Management: Implementing AI-Generated Avatars for Industrial Process Monitoring
Traditional fault notification systems in industrial environments depend on conventional technology like stack lights and auditory alarms. These systems provide limited contextual information and require prior operator training for accurate operation interpretation. This paper proposes an avatar-based fault management system that enhances human-machine interaction by replacing conventional alarm mechanisms with AI-generated avatars delivering structured, real-time notifications. The system is implemented on a Raspberry Pi that processes sensor inputs, classifies detected anomalies, and selects the appropriate pre-recorded video message stored on an SD card. These videos feature realistic avatars designed to avoid the uncanny valley effect, ensuring effective and intu- itive communication.
- 6 October 2025
- Published by Astra Ediciones
- p. 213-223
Humanizing Fault Management: Implementing AI-Generated Avatars for Industrial Process Monitoring
Traditional fault notification systems in industrial environments depend on conventional technology like stack lights and auditory alarms. These systems provide limited contextual information and require prior operator training for accurate operation interpretation. This paper proposes an avatar-based fault management system that enhances human-machine interaction by replacing conventional alarm mechanisms with AI-generated avatars delivering structured, real-time notifications. The system is implemented on a Raspberry Pi that processes sensor inputs, classifies detected anomalies, and selects the appropriate pre-recorded video message stored on an SD card. These videos feature realistic avatars designed to avoid the uncanny valley effect, ensuring effective and intu- itive communication.
- 6 October 2025
- Published by Astra Ediciones
- p. 133-143
Integration of Generative AI into a Project-based Learning Activity in a Microcontrollers Course
The use of generative artificial intelligence by students has increased since the launch of powerful natural language processing tools such as ChatGPT. Several works have highlighted the strengths, weaknesses, opportunities, and threats of this technology in education, but there is a reduced amount of empirical research of its application in real academic scenarios. This work proposes the use of ChatGPT as a tool that could help engineering students solve highly complex problems more efficiently and accurately. In order to test this hypothesis, an experience with a group of students from a mechatronics engineering microcontrollers course was carried out using the project-based engineering approach. The results highlight the necessity of finding a balance between a learner’s previous knowledge and the complexity of a new task and educating students about proper prompt engineering techniques and the use of ChatGPT and any other GAI only as a complement to their own knowledge and skills.
- 6 October 2025
- Published by Astra Ediciones
- p. 133-143
Integration of Generative AI into a Project-based Learning Activity in a Microcontrollers Course
The use of generative artificial intelligence by students has increased since the launch of powerful natural language processing tools such as ChatGPT. Several works have highlighted the strengths, weaknesses, opportunities, and threats of this technology in education, but there is a reduced amount of empirical research of its application in real academic scenarios. This work proposes the use of ChatGPT as a tool that could help engineering students solve highly complex problems more efficiently and accurately. In order to test this hypothesis, an experience with a group of students from a mechatronics engineering microcontrollers course was carried out using the project-based engineering approach. The results highlight the necessity of finding a balance between a learner’s previous knowledge and the complexity of a new task and educating students about proper prompt engineering techniques and the use of ChatGPT and any other GAI only as a complement to their own knowledge and skills.
- Open Access
Thermal Field Reconstruction on Microcontrollers: A Physics-Informed Digital Twin Using Laplace Equation and Real-Time Sensor Data
This paper presents a physics-informed digital twin designed for real-time thermal monitoring and visualization of a metallic plate. The system comprises a physical layer consisting of an aluminum plate equipped with thermistors to capture boundary conditions, a computational layer that implements the steady-state Laplace equation using the finite difference method, and an embedded execution framework deployed on a microcontroller that utilizes Direct Memory Access-driven ADC for efficient concurrent acquisition. The computed thermal field is transmitted through a serial interface and displayed in real time using a Python-based visualization interface. The Steinhart–Hart model was used to experimentally characterize the sensors, ensuring accuracy in the boundary condition acquisition. While the current formulation is restricted to steady-state conditions, it enables accurate spatial reconstructions with acceptable error margins and demonstrates operational concurrency with the physical system. The compact and modular architecture allows adaptation to other physical domains governed by elliptic PDEs, making it suitable for educational applications, diagnostic prototyping, and embedded edge deployments.
Open AccessThermal Field Reconstruction on Microcontrollers: A Physics-Informed Digital Twin Using Laplace Equation and Real-Time Sensor Data
This paper presents a physics-informed digital twin designed for real-time thermal monitoring and visualization of a metallic plate. The system comprises a physical layer consisting of an aluminum plate equipped with thermistors to capture boundary conditions, a computational layer that implements the steady-state Laplace equation using the finite difference method, and an embedded execution framework deployed on a microcontroller that utilizes Direct Memory Access-driven ADC for efficient concurrent acquisition. The computed thermal field is transmitted through a serial interface and displayed in real time using a Python-based visualization interface. The Steinhart–Hart model was used to experimentally characterize the sensors, ensuring accuracy in the boundary condition acquisition. While the current formulation is restricted to steady-state conditions, it enables accurate spatial reconstructions with acceptable error margins and demonstrates operational concurrency with the physical system. The compact and modular architecture allows adaptation to other physical domains governed by elliptic PDEs, making it suitable for educational applications, diagnostic prototyping, and embedded edge deployments.
- Open Access
- 11 July 2025
- Published by MDPI AG in Preprints.org
Thermal Field Reconstruction on Microcontrollers: A Physics-Informed Digital Twin Using Laplace Equation and Real-Time Sensor Data
This paper presents a physics-informed digital twin designed for real-time thermal monitoring and visualization of a metallic plate. The system comprises a physical layer consisting of an aluminum plate equipped with thermistors to capture boundary conditions, a computational layer that implements the steady-state Laplace equation using the finite difference method, and an embedded execution framework deployed on a microcontroller that utilizes Direct Memory Access-driven ADC for efficient concurrent acquisition. The computed thermal field is transmitted through a serial interface and displayed in real time using a Python-based visualization interface. The Steinhart–Hart model was used to experimentally characterize the sensors, ensuring accuracy in the boundary condition acquisition. Results demonstrate that the proposed architecture can simulate and reconstruct internal thermal states with high spatial resolution in resource-constrained hardware. This affordable, real-time approach is suitable for educational applications, industrial training, industrial thermal diagnostics, and the prototyping of embedded digital twin systems. Future extensions include adaptive mesh refinement, dynamic thermal modeling, and integration with control systems.
- 11 July 2025
- Published by MDPI AG in Preprints.org
Open AccessThermal Field Reconstruction on Microcontrollers: A Physics-Informed Digital Twin Using Laplace Equation and Real-Time Sensor Data
This paper presents a physics-informed digital twin designed for real-time thermal monitoring and visualization of a metallic plate. The system comprises a physical layer consisting of an aluminum plate equipped with thermistors to capture boundary conditions, a computational layer that implements the steady-state Laplace equation using the finite difference method, and an embedded execution framework deployed on a microcontroller that utilizes Direct Memory Access-driven ADC for efficient concurrent acquisition. The computed thermal field is transmitted through a serial interface and displayed in real time using a Python-based visualization interface. The Steinhart–Hart model was used to experimentally characterize the sensors, ensuring accuracy in the boundary condition acquisition. Results demonstrate that the proposed architecture can simulate and reconstruct internal thermal states with high spatial resolution in resource-constrained hardware. This affordable, real-time approach is suitable for educational applications, industrial training, industrial thermal diagnostics, and the prototyping of embedded digital twin systems. Future extensions include adaptive mesh refinement, dynamic thermal modeling, and integration with control systems.
- Open Access
- 26 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 38181-38194
Lightweight Deep Learning for sEMG-Based Fingers Position Classification and Embedded System Deployment
This paper presents a lightweight deep learning approach for classifying and tracking finger positions based on surface electromyography (sEMG) signals, designed with a perspective of its implementation in embedded systems. Unlike traditional studies that focus on static hand gestures, this research highlights the continuous tracking of finger positions during dynamic hand movements, such as the transition between open and closed states. A 1D Convolutional Neural Network was developed and validated using experimental data, achieving a classification accuracy of 97% in controlled scenarios. The architecture of the model balances computational efficiency and classification performance, making it deployable on resource-constrained embedded technology. This feature highlights its potential for real-time applications in prosthetics, robotics, and human-computer interaction. Although further optimization is needed for better generalization to unseen data, this study emphasizes the significance of developing deployable algorithms that excel beyond simulation environments focusing on enhancing model robustness and validating its real-time performance through hardware-based implementations. The findings indicate notable advancements in connecting sophisticated machine-learning techniques with effective embedded solutions for complex, dynamic tasks.
- 26 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 38181-38194
Open AccessLightweight Deep Learning for sEMG-Based Fingers Position Classification and Embedded System Deployment
This paper presents a lightweight deep learning approach for classifying and tracking finger positions based on surface electromyography (sEMG) signals, designed with a perspective of its implementation in embedded systems. Unlike traditional studies that focus on static hand gestures, this research highlights the continuous tracking of finger positions during dynamic hand movements, such as the transition between open and closed states. A 1D Convolutional Neural Network was developed and validated using experimental data, achieving a classification accuracy of 97% in controlled scenarios. The architecture of the model balances computational efficiency and classification performance, making it deployable on resource-constrained embedded technology. This feature highlights its potential for real-time applications in prosthetics, robotics, and human-computer interaction. Although further optimization is needed for better generalization to unseen data, this study emphasizes the significance of developing deployable algorithms that excel beyond simulation environments focusing on enhancing model robustness and validating its real-time performance through hardware-based implementations. The findings indicate notable advancements in connecting sophisticated machine-learning techniques with effective embedded solutions for complex, dynamic tasks.
- Open Access
Wavelet-Based Computational Intelligence for Real-Time Anomaly Detection and Fault Isolation in Embedded Systems
In today’s technologically advanced landscape, sensors feed critical data for accurate decision-making and actions. Ensuring the integrity and reliability of sensor data is paramount to system performance and security. This paper introduces an innovative approach utilizing discrete wavelet transforms (DWT) embedded within microcontrollers to scrutinize sensor data meticulously. Our methodology aims to detect and isolate malfunctions, misuse, or any anomalies before they permeate the system, potentially causing widespread disruption. By leveraging the power of wavelet-based analysis, we embed computational intelligence directly into the microcontrollers, enabling them to monitor and validate their outputs in real-time. This proactive anomaly detection framework is designed to distinguish between normal and aberrant sensor behaviors, thereby safeguarding the system from erroneous data propagation. Our approach significantly enhances the reliability of embedded systems, providing a robust defense against false data injection attacks and contributing to overall cybersecurity.
Open AccessWavelet-Based Computational Intelligence for Real-Time Anomaly Detection and Fault Isolation in Embedded Systems
In today’s technologically advanced landscape, sensors feed critical data for accurate decision-making and actions. Ensuring the integrity and reliability of sensor data is paramount to system performance and security. This paper introduces an innovative approach utilizing discrete wavelet transforms (DWT) embedded within microcontrollers to scrutinize sensor data meticulously. Our methodology aims to detect and isolate malfunctions, misuse, or any anomalies before they permeate the system, potentially causing widespread disruption. By leveraging the power of wavelet-based analysis, we embed computational intelligence directly into the microcontrollers, enabling them to monitor and validate their outputs in real-time. This proactive anomaly detection framework is designed to distinguish between normal and aberrant sensor behaviors, thereby safeguarding the system from erroneous data propagation. Our approach significantly enhances the reliability of embedded systems, providing a robust defense against false data injection attacks and contributing to overall cybersecurity.
- Open Access
Estimation of the Availability of Electrical Energy from the Thermal Energy Extracted by Thermoelectric Modules: Case Study in Monterrey, México
This research proposes the utilization of Peltier modules to convert electrical energy from thermal energy to show its potential as a renewable energy source for residential and commercial application. The study, whose results are presented in this manuscript, was conducted in the city of Monterrey, located in the northeast of Mexico. The energy source was tested and analyzed utilizing a set of statistical metrics and further comparison against experimental test results. The Distrito Tec area in Monterrey city is an academic complex and it was chosen for this study, which consisted of the indirect measurement of the average annual heat energy stored within the buildings’ cement structure. The aim was to obtain the annual accumulated electrical power in Wh per year that Peltier modules could provide in Distrito Tec, which is located in a city with solar irradiation levels above the world average. The proposal in this paper could encourage further investigation regarding this energy that is currently waste heat. More specifically, the results of this research highlight the importance of thermoelectric modules and seek to motivate research to improve their properties and make them more efficient and more viable as well. Thermoelectric modules have the potential to be part of the solution to sustainable development as presented in the United Nations SDG-7—ensure access to affordable, reliable, sustainable and modern energy for all.
Open AccessEstimation of the Availability of Electrical Energy from the Thermal Energy Extracted by Thermoelectric Modules: Case Study in Monterrey, México
This research proposes the utilization of Peltier modules to convert electrical energy from thermal energy to show its potential as a renewable energy source for residential and commercial application. The study, whose results are presented in this manuscript, was conducted in the city of Monterrey, located in the northeast of Mexico. The energy source was tested and analyzed utilizing a set of statistical metrics and further comparison against experimental test results. The Distrito Tec area in Monterrey city is an academic complex and it was chosen for this study, which consisted of the indirect measurement of the average annual heat energy stored within the buildings’ cement structure. The aim was to obtain the annual accumulated electrical power in Wh per year that Peltier modules could provide in Distrito Tec, which is located in a city with solar irradiation levels above the world average. The proposal in this paper could encourage further investigation regarding this energy that is currently waste heat. More specifically, the results of this research highlight the importance of thermoelectric modules and seek to motivate research to improve their properties and make them more efficient and more viable as well. Thermoelectric modules have the potential to be part of the solution to sustainable development as presented in the United Nations SDG-7—ensure access to affordable, reliable, sustainable and modern energy for all.