Frederic Domingue
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Frederic Domingue has a total of 18 co-authors. Here are the top co-authors by number of shared publications:
- Guy Ayissi Eyebe
Department of Electrical and Computer Engineering, Université du Québec à Trois-Rivières, Trois-Rivières, QC G9A 5H7, Canada
2 shared publications - Abdelhak Hamid Allah
Department of Electrical and Computer Engineering, Université du Québec à Trois-Rivières, Trois-Rivières, QC G9A 5H7, Canada
1 shared publication - Benoit Bideau
Institut d'Innovations en Écomatériaux, Écoproduits et Écoénergies, Université du Québec à Trois-Riviàres
1 shared publication - Carl Rivkin
Transportation and Hydrogen Systems Center, National Renewable Energy Laboratory, 15013 Denver West Parkway, Golden, CO 80401-3305, USA
1 shared publication - David Myja
Institut d’Innovations en Écomatériaux, Écoproduits et Écoénergies à base de biomasse, Université du Québec à Trois-Rivières, P.O. Box 500, Trois-Rivières, Québec Canada, G9A 5H7
1 shared publication - Eric Loranger
Université du Québec à Trois-Rivières
1 shared publication
Latest Publications
- 28 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Embedded Systems Letters
- Vol. 18 (2) , 115-118
A Memory Representation of Random Forests Optimized for Resource-Limited Embedded Devices
Random forests are a versatile and effective machine learning technique widely applied across various tasks. With the increasing demand for deploying machine learning models on resource-constrained embedded devices, such as microcontrollers, challenges arise from the growing complexity of modern datasets. These challenges often result in models that are too large in memory and storage requirements to be feasibly implemented on small devices. In this work, we propose a lossless memory representation of random forests that significantly limits the amount of random-access memory (RAM) required for prediction tasks, while also reducing the amount of non-volatile memory needed to store the model. The approach achieves efficiency by embedding the data of leaf nodes within the decision nodes, thereby streamlining the tree structure. Additionnally, it supports in-place prediction without requiring a decompression step. To evaluate our method, we implemented four random forests derived from real-world datasets onto four microcontroller platforms. Our results demonstrate that prediction tasks can be performed using at most 144 bytes of RAM for classification tasks, and at most 48 bytes for regression tasks, while memory accesses account for a maximum of 27.0% of the total CPU cycles. On the fastest platform, prediction times ranged between 59 and 75 μs, highlighting the suitability of this method for a variety of real-time applications.
- 28 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Embedded Systems Letters
- Vol. 18 (2) , 115-118
A Memory Representation of Random Forests Optimized for Resource-Limited Embedded Devices
Random forests are a versatile and effective machine learning technique widely applied across various tasks. With the increasing demand for deploying machine learning models on resource-constrained embedded devices, such as microcontrollers, challenges arise from the growing complexity of modern datasets. These challenges often result in models that are too large in memory and storage requirements to be feasibly implemented on small devices. In this work, we propose a lossless memory representation of random forests that significantly limits the amount of random-access memory (RAM) required for prediction tasks, while also reducing the amount of non-volatile memory needed to store the model. The approach achieves efficiency by embedding the data of leaf nodes within the decision nodes, thereby streamlining the tree structure. Additionnally, it supports in-place prediction without requiring a decompression step. To evaluate our method, we implemented four random forests derived from real-world datasets onto four microcontroller platforms. Our results demonstrate that prediction tasks can be performed using at most 144 bytes of RAM for classification tasks, and at most 48 bytes for regression tasks, while memory accesses account for a maximum of 27.0% of the total CPU cycles. On the fastest platform, prediction times ranged between 59 and 75 μs, highlighting the suitability of this method for a variety of real-time applications.
- Open Access
Improved Fully 3D-Printed SIW-Based Sensor for Non-Invasive Glucose Measurement
This paper presents a fully 3D-printed microfluidic microwave sensor based on substrate-integrated waveguide (SIW) technology for detecting glucose levels in liquid, aimed at monitoring diabetic patients. The sensor’s design features a circular SIW cavity with an integrated sample holder placed in the cavity’s center, maximizing the electric field disturbance from the liquid under test (LUT). Operating in the TM010 mode at a resonance frequency around 5.740 GHz, the sensor detects glucose concentrations by measuring resonance frequency shifts in the S11 reflection response across glucose levels. A conductive sheath partially covers the sample holder to enhance sensitivity and improve the limit of detection (LOD) by increasing field penetration into the LUT. Fabricated using an additive manufacturing electronics (AMEs) method, the sensor is produced in a single pass without post-processing. The experimental validation confirms its high sensitivity of 1.218 MHz/(mg/dL) and a low limit of detection of 0.774 mg/dL in the glucose concentration range (10–200 mg/dL), reflecting typical Type 2 diabetes levels. The key advantages of the sensor include its compactness, enhanced sensitivity and limit of detection, innovative manufacturing, and cost-effectiveness, supporting its potential as a non-invasive glucose monitoring tool. This study establishes a proof of concept for the in vitro measurement of glucose, demonstrating the sensor’s ability to provide accurate and reliable results in a controlled environment.
Open AccessImproved Fully 3D-Printed SIW-Based Sensor for Non-Invasive Glucose Measurement
This paper presents a fully 3D-printed microfluidic microwave sensor based on substrate-integrated waveguide (SIW) technology for detecting glucose levels in liquid, aimed at monitoring diabetic patients. The sensor’s design features a circular SIW cavity with an integrated sample holder placed in the cavity’s center, maximizing the electric field disturbance from the liquid under test (LUT). Operating in the TM010 mode at a resonance frequency around 5.740 GHz, the sensor detects glucose concentrations by measuring resonance frequency shifts in the S11 reflection response across glucose levels. A conductive sheath partially covers the sample holder to enhance sensitivity and improve the limit of detection (LOD) by increasing field penetration into the LUT. Fabricated using an additive manufacturing electronics (AMEs) method, the sensor is produced in a single pass without post-processing. The experimental validation confirms its high sensitivity of 1.218 MHz/(mg/dL) and a low limit of detection of 0.774 mg/dL in the glucose concentration range (10–200 mg/dL), reflecting typical Type 2 diabetes levels. The key advantages of the sensor include its compactness, enhanced sensitivity and limit of detection, innovative manufacturing, and cost-effectiveness, supporting its potential as a non-invasive glucose monitoring tool. This study establishes a proof of concept for the in vitro measurement of glucose, demonstrating the sensor’s ability to provide accurate and reliable results in a controlled environment.
- 13 June 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Letters
- Vol. 3 (7) , 1-4
Coplanar waveguide techno TEMPO Oxidized Thermomechanical Pulp-Based Microwave Humidity Sensor
The thermomechanical pulp (TMP) is a product of lignocellulosic biomass. This paper investigates TEMPO oxidized TMP (TO-TMP) as an ecofriendly, low-cost and highly sensitive dielectric material for humidity sensing applications. The TEMPO oxidation was used to increase the water absorption of TMP while overcoming the adverse hydropho-bic effects imparted to lignin. The physical and the dielectric effects of TEMPO oxidation on a TMP sheet having a carboxyl rate of 80 mmol/kg were studied. An experimental validation with an original sensing scheme in frequency shift paradigm involving a microwave resonator in coplanar waveguide (CPW) technology was proposed. In the $52 − 76.2$ %RH range, the sensitivity rose from 7.76 MHz/%R to 9.04 MHz/%RH after the TEMPO oxidation increased the carboxyl content to 1164 mmol/kg, and to 12.1 MHz/%RH when the carboxyl content was enhanced up to 2233 mmol/kg. The effects of the TEMPO oxidation degree and the thickness variation are also studied.- 13 June 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Letters
- Vol. 3 (7) , 1-4
Coplanar waveguide techno TEMPO Oxidized Thermomechanical Pulp-Based Microwave Humidity Sensor
The thermomechanical pulp (TMP) is a product of lignocellulosic biomass. This paper investigates TEMPO oxidized TMP (TO-TMP) as an ecofriendly, low-cost and highly sensitive dielectric material for humidity sensing applications. The TEMPO oxidation was used to increase the water absorption of TMP while overcoming the adverse hydropho-bic effects imparted to lignin. The physical and the dielectric effects of TEMPO oxidation on a TMP sheet having a carboxyl rate of 80 mmol/kg were studied. An experimental validation with an original sensing scheme in frequency shift paradigm involving a microwave resonator in coplanar waveguide (CPW) technology was proposed. In the $52 − 76.2$ %RH range, the sensitivity rose from 7.76 MHz/%R to 9.04 MHz/%RH after the TEMPO oxidation increased the carboxyl content to 1164 mmol/kg, and to 12.1 MHz/%RH when the carboxyl content was enhanced up to 2233 mmol/kg. The effects of the TEMPO oxidation degree and the thickness variation are also studied.- 1 February 2014
- Published by Elsevier in International Journal of Hydrogen Energy
- Vol. 39 (9) , 4664-4673
Assessment of commercial micro-machined hydrogen sensors performance metrics for safety sensing applications
- 1 February 2014
- Published by Elsevier in International Journal of Hydrogen Energy
- Vol. 39 (9) , 4664-4673
Assessment of commercial micro-machined hydrogen sensors performance metrics for safety sensing applications
- 15 March 2010
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Microwave Theory and Techniques
- Vol. 58 (4) , 1056-1064
Distributed MEMS Tunable Impedance-Matching Network Based on Suspended Slow-Wave Structure Fabricated in a Standard CMOS Technology
A tunable RF microelectromechanical system (MEMS) impedance-matching network operating at a frequency band from 13 to 24 GHz based on the distributed microelectromechanical transmission line (DMTL) concept is presented in this paper. The network is implemented using a standard 0.35- ¿m CMOS technology and employs a novel suspended slow-wave (SSW) structure on a silicon substrate. The SSW structure results in a reduced total footprint and enhanced impedance coverage. The 8-bit DMTL matching network, fabricated using switched MEMS capacitors and SSW coplanar waveguide on a silicon substrate, results in a wide coverage of the Smith chart up to a maximum voltage standing-wave ratio of 11.5:1 with an impedance matching better than 10 dB and a power transfer ratio of better than -2.84 dB at 24 GHz. To our knowledge, this is the first implementation of a DMTL tunable MEMS impedance-matching network using a standard CMOS technology.
- 15 March 2010
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Microwave Theory and Techniques
- Vol. 58 (4) , 1056-1064
Distributed MEMS Tunable Impedance-Matching Network Based on Suspended Slow-Wave Structure Fabricated in a Standard CMOS Technology
A tunable RF microelectromechanical system (MEMS) impedance-matching network operating at a frequency band from 13 to 24 GHz based on the distributed microelectromechanical transmission line (DMTL) concept is presented in this paper. The network is implemented using a standard 0.35- ¿m CMOS technology and employs a novel suspended slow-wave (SSW) structure on a silicon substrate. The SSW structure results in a reduced total footprint and enhanced impedance coverage. The 8-bit DMTL matching network, fabricated using switched MEMS capacitors and SSW coplanar waveguide on a silicon substrate, results in a wide coverage of the Smith chart up to a maximum voltage standing-wave ratio of 11.5:1 with an impedance matching better than 10 dB and a power transfer ratio of better than -2.84 dB at 24 GHz. To our knowledge, this is the first implementation of a DMTL tunable MEMS impedance-matching network using a standard CMOS technology.