Comparative Analysis of Machine Learning Techniques for Temperature Compensation in Microwave Sensors

Kazemi, Nazli; Abdolrazzaghi, Mohammad; Musilek, Petr

doi:10.1109/tmtt.2021.3081119

Cited by 81 publications

(32 citation statements)

References 50 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Sensor calibration and cross-sensitivity compensation are receiving growing attention in the wider field of ML [33]- [35], [38], [39], [42]- [45]. In this section, we position our work based on Bayesian learning, regression, and design of experiment within the context of these studies, which have relied on support vector machines (SVMs) [34], random forests (RFs) [34], [38], [39], Gaussian process regression (GPR) [39], [42], [43], and artificial neural networks (ANNs) taking most often the form of multilayered perceptrons (MLP) [33], [34], [44], [45] and fuzzy neural networks (FNNs) [35], among other methodologies [34].…”

Section: B Comparison With Other ML Methodsmentioning

confidence: 99%

“…Besides polynomial regression, a variety of machine learning (ML) methods have been successfully used to calibrate liquid mixture sensors [33]- [35], gas sensors [36]- [40], pH sensors [41], thermal and differential-pressure anemometers [42], a tactile sensor [43], and a photonic sensor [44]. Some of these sensors were successfully compensated against the influence of temperature [33], [34], [38], [45], relative humidity [38], [45], and chemical cross-sensitivities [38].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Bayesian Sensor Calibration

Berger

Schott²,

Oliver

2022

IEEE Sensors J.

View full text Add to dashboard Cite

The calibration of multisensor systems can cause significant costs in terms of time and resources, in particular when crosssensitivities to parasitic influences are to be compensated. Successful calibration ensures the trustworthy subsequent operation of a sensor system, guaranteeing that one or several measurands of interest can be inferred from its output signals with specified uncertainty. As the present study shows, this goal can be reached by reduced calibration procedures with fewer calibration conditions than parameters needed to model the device response. This is achieved using Bayesian inference by combining calibration data of a sensor system with statistical prior information about the ensemble to which it belongs. Optimal reduced sets of calibration conditions are identified by the method of Bayesian experimental design. The method is demonstrated on a Hall-temperature sensor system whose nonlinear response model requires seven parameters in the temperature range between -30 °C and 150 °C and for magnetic field values B between -25 mT and 25 mT. For the prior, a multivariate normal distribution of the model parameters is acquired using 14 specimens of the sensor ensemble. I-optimal calibration at one, two, and three temperatures reduces the rms standard deviation of B inferred from sensor output signals from 203 µT before calibration down to 78 µT, 41 µT, and 34 µT, respectively. Similar conclusions apply to G-optimal calibration. The paper describes how to implement the Bayesian prior acquisition, inference, and experimental design. The proposed approach can help save resources and cut costs in sensor calibration.

show abstract

Section: B Comparison With Other ML Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Bayesian Sensor Calibration

Berger

Schott²,

Oliver

2022

IEEE Sensors J.

View full text Add to dashboard Cite

show abstract

“…In particular, the frequency response characteristic of the filter is changed by a change in the capacitance value related to coupling. Therefore, biosensor applications are possible because the filter can detect specific substances on tissues [21].…”

Section: Resultsmentioning

confidence: 99%

Miniaturized Dual-Band Bandpass Filter Using T-Shaped Line Based on Stepped Impedance Resonator with Meander Line and Folded Structure

2022

View full text Add to dashboard Cite

A stepped impedance resonator (SIR) is suitable for designing a dual-band bandpass filter (BPF) that can be adjusted to reject spurious bands. A BPF is proposed using an SIR T-shaped meander line and folded structure. The BPF mainly comprises a meander line, a folded structure, and a T-shaped line. A novel BPF is used for the T-shaped line, which operates as a band-stop filter connecting to the center of the BPF. As a result, the complete BPF enables dual-band operation. The insertion and return losses of the first frequency passband (f01) are 0.024 and 17.3 dB, respectively, with a bandwidth of 46% at a center frequency of 2.801 GHz (2.2–3.48 GHz). The insertion and return losses of the second frequency passband (f02) are 0.026 and 17.2 dB, respectively, with a bandwidth of 10% at a center frequency of 4.351 GHz (4.13–4.55 GHz). The proposed BPF provides low loss, a simple structure, and a small size of only 4.29 × 4.08 mm, and it can be integrated into mobile communications systems.

show abstract

“…In addition, the fourth industrial revolution, big data, autonomous driving functions, and sensors are increasingly used. In the case of a bandpass filter, it is sometimes applied as a sensor that detects a resonance phenomenon and passes a desired frequency band [ 3 , 4 ]. The sensor–based bandpass filter is also applied to medical systems for diagnosis or treatment [ 3 , 4 ].…”

Section: Introductionmentioning

confidence: 99%

“…In the case of a bandpass filter, it is sometimes applied as a sensor that detects a resonance phenomenon and passes a desired frequency band [ 3 , 4 ]. The sensor–based bandpass filter is also applied to medical systems for diagnosis or treatment [ 3 , 4 ]. Among BPFs, stub BPFs are known for having a simple structure and being easy to produce.…”

Section: Introductionmentioning

confidence: 99%

Miniaturized Bandpass Filter Using a Combination of T–Shaped Folded SIR Short Loaded Stubs

Yoon

Kim

Lee

2022

Sensors

View full text Add to dashboard Cite

The consumption of multimedia content is ubiquitous in modern society. This is made possible by wireless local area networks (W–LAN) or wire service systems. Bandpass filters (BPF) have become very popular as they solve certain data transmission limitations allowing users to obtain reliable access to their multimedia content. The BPFs with quarter–wavelength short stubs can achieve performance; however, these BPFs are bulky. In this article, we propose a compact BPF with a T–shaped stepped impedance resonator (SIR) transmission line and a folded SIR structure. The proposed BPF uses a T–shaped SIR connected to a J–inverter structure (transmission line); this T–shaped SIR structure is used to replace the λg/4 transmission line seen in conventional stub BPFs. In addition, a folded SIR is added to the short stubs seen in conventional stub BPFs. This approach allows us to significantly reduce the size of the BPF. The advantage of a BPF is its very small size, low insertion loss, and wide bandwidth. The overall size of the new BPF is 2.44 mm × 1.49 mm (0.068λg × 0.059λg). The proposed BPF can be mass produced using semiconductors due to its planar structure. This design has the potential to be widely used in various areas including military, medical, and industrial systems.

show abstract

Comparative Analysis of Machine Learning Techniques for Temperature Compensation in Microwave Sensors

Cited by 81 publications

References 50 publications

Bayesian Sensor Calibration

Bayesian Sensor Calibration

Miniaturized Dual-Band Bandpass Filter Using T-Shaped Line Based on Stepped Impedance Resonator with Meander Line and Folded Structure

Miniaturized Bandpass Filter Using a Combination of T–Shaped Folded SIR Short Loaded Stubs

Contact Info

Product

Resources

About