Pub Date : 2025-11-24DOI: 10.1109/LSENS.2025.3636563
Anyu Li;Songping Mai
Human activity recognition (HAR) is critical in applications, such as medical surveillance and contactless human–machine interaction. This letter proposes an accurate and lightweight 1-D convolutional neural network for radar point-cloud-based HAR. The model comprises a lightweight Ghost-PointNet to extract spatial features directly from raw point clouds and a multibranch network to capture multiscale temporal dependencies across the extracted features for classification. Evaluated on the MMActivity dataset, our model achieves 97.82% test accuracy with only 59k parameters, 1.37 MB peak runtime memory usage, and 0.23 giga floating-point operations (GFLOPs). Compared to methods achieving similar accuracy, ours requires up to 99% fewer parameters and 96% smaller input data volumes. By integrating model simplification, ghost convolutions, and depthwise separable convolutions, we maintain high accuracy while drastically reducing computational costs. The ultra-lightweight model is ideal for deployment on resource-constrained edge devices.
{"title":"Accurate and Efficient 1D-CNN for Human Activity Recognition Using Millimeter-Wave Point Clouds","authors":"Anyu Li;Songping Mai","doi":"10.1109/LSENS.2025.3636563","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3636563","url":null,"abstract":"Human activity recognition (HAR) is critical in applications, such as medical surveillance and contactless human–machine interaction. This letter proposes an accurate and lightweight 1-D convolutional neural network for radar point-cloud-based HAR. The model comprises a lightweight Ghost-PointNet to extract spatial features directly from raw point clouds and a multibranch network to capture multiscale temporal dependencies across the extracted features for classification. Evaluated on the MMActivity dataset, our model achieves 97.82% test accuracy with only 59k parameters, 1.37 MB peak runtime memory usage, and 0.23 giga floating-point operations (GFLOPs). Compared to methods achieving similar accuracy, ours requires up to 99% fewer parameters and 96% smaller input data volumes. By integrating model simplification, ghost convolutions, and depthwise separable convolutions, we maintain high accuracy while drastically reducing computational costs. The ultra-lightweight model is ideal for deployment on resource-constrained edge devices.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"10 1","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-11-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778371","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-24DOI: 10.1109/LSENS.2025.3636392
Si Young Choi;Cierra Anderson;Kevin Dai;Ilaria Fratelli;Yi-Chen Liu;Peter Ballentine;Yuewen Tan;Michael Bardash;Guy Garty;Andrew Harken;Ioannis Kymissis;Savannah Eisner
This letter reports the first characterization of InAlN/gallium nitride (GaN)-on-Si solid-state detectors under ultra-high dose-per-pulse (DPP) electron irradiation for FLASH radiotherapy applications. Unpassivated Ni/Au metal-semiconductor-metal and 2-D electron gas (2DEG) interdigitated transducer (IDT) detectors exhibited the best transient performance, with the 2DEG IDT resolving individual pulses as short as 3 μs rise and 325 μs fall times and achieving a signal-to-noise ratio exceeding 220 under 2 Gy pulses. Across a DPP range of 0.05–0.5 Gy, the 2DEG IDT detector demonstrated consistent, dose-dependent voltage responses with sensitivities of 0.5–0.9 V/Gy and a normalized sensitivity of 48.75 nC/Gy/mm$^{2}$, one to two orders of magnitude higher than reported silicon carbide (SiC) and diamond detectors. These results establish InAlN/GaN detectors as promising candidates for compact, high-speed, and radiation-tolerant dosimetry systems capable of real-time single-pulse monitoring in clinical FLASH radiotherapy.
{"title":"InAlN/GaN Detectors for Real-Time Dosimetry in Ultra-High Dose Rate FLASH Radiotherapy","authors":"Si Young Choi;Cierra Anderson;Kevin Dai;Ilaria Fratelli;Yi-Chen Liu;Peter Ballentine;Yuewen Tan;Michael Bardash;Guy Garty;Andrew Harken;Ioannis Kymissis;Savannah Eisner","doi":"10.1109/LSENS.2025.3636392","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3636392","url":null,"abstract":"This letter reports the first characterization of InAlN/gallium nitride (GaN)-on-Si solid-state detectors under ultra-high dose-per-pulse (DPP) electron irradiation for FLASH radiotherapy applications. Unpassivated Ni/Au metal-semiconductor-metal and 2-D electron gas (2DEG) interdigitated transducer (IDT) detectors exhibited the best transient performance, with the 2DEG IDT resolving individual pulses as short as 3 μs rise and 325 μs fall times and achieving a signal-to-noise ratio exceeding 220 under 2 Gy pulses. Across a DPP range of 0.05–0.5 Gy, the 2DEG IDT detector demonstrated consistent, dose-dependent voltage responses with sensitivities of 0.5–0.9 V/Gy and a normalized sensitivity of 48.75 nC/Gy/mm<inline-formula><tex-math>$^{2}$</tex-math></inline-formula>, one to two orders of magnitude higher than reported silicon carbide (SiC) and diamond detectors. These results establish InAlN/GaN detectors as promising candidates for compact, high-speed, and radiation-tolerant dosimetry systems capable of real-time single-pulse monitoring in clinical FLASH radiotherapy.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"10 1","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-11-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145705931","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-24DOI: 10.1109/LSENS.2025.3636379
Ashi Singhal;Fatima Sami;Adeeba Khan;Mohd Wajid;Mohammed Usman;Hemant Kumar Meena;Abhishek Srivastava
This study presents a novel sensor-based system for human vital sign estimation through a glass barrier using Texas Instruments 77 GHz AWR1843BOOST frequency modulated continuous wave radar. To improve signal fidelity and ensure precise extraction of physiological parameters, the signal processing pipeline incorporates beamforming, variational mode decomposition, and recursive Bayesian estimation. Experiments were conducted across distances of 0.4–0.7 m through glass barrier. The system achieved an mean absolute error of 1.93 breaths per minute for respiratory rate and 8.47 beats per minute for heart rate, with average accuracies of 83.74% and 91.41%, respectively. The results highlight the system’s potential for reliable, hygienic, noninvasive monitoring in sensitive settings, such as neonatal intensive care units and infection control wards, where contact-based methods are risky.
{"title":"Remote Vital Sign Monitoring Through Glass Barrier Using 77GHz FMCW Radar","authors":"Ashi Singhal;Fatima Sami;Adeeba Khan;Mohd Wajid;Mohammed Usman;Hemant Kumar Meena;Abhishek Srivastava","doi":"10.1109/LSENS.2025.3636379","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3636379","url":null,"abstract":"This study presents a novel sensor-based system for human vital sign estimation through a glass barrier using Texas Instruments 77 GHz AWR1843BOOST frequency modulated continuous wave radar. To improve signal fidelity and ensure precise extraction of physiological parameters, the signal processing pipeline incorporates beamforming, variational mode decomposition, and recursive Bayesian estimation. Experiments were conducted across distances of 0.4–0.7 m through glass barrier. The system achieved an mean absolute error of 1.93 breaths per minute for respiratory rate and 8.47 beats per minute for heart rate, with average accuracies of 83.74% and 91.41%, respectively. The results highlight the system’s potential for reliable, hygienic, noninvasive monitoring in sensitive settings, such as neonatal intensive care units and infection control wards, where contact-based methods are risky.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"10 1","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-11-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778197","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
We developed polymer-coated resonant sensors to monitor volatile fatty acid (VFA) levels in the cow rumen. A 19-nm polybutadiene (PBD) film was applied to a quartz crystal microbalance (QCM), causing a frequency shift of approximately −400 Hz at 1 mM for all VFAs. As VFA concentrations increase, the frequency shift also increases in magnitude, highlighting the sensor's high sensitivity to VFAs. To test selectivity, a 171-nm-thick polydimethylsiloxane-coated QCM sensor was used. The frequency shifts observed were −178 Hz for acetic acid, −205 Hz for propionic acid, and −78 Hz for butyric acid, indicating different adsorption behaviors. Subsequently, we developed a VFA resonant sensor capsule for in vivo rumen monitoring. This capsule includes a PBD-coated QCM, a temperature/pressure sensor, an oscillator, a microcontroller, and a battery. In vivo tests showed a significant drop in resonant frequency within the cow rumen, demonstrating the detection of overall VFAs. The nano-thin polymer film coating on the resonant sensor demonstrates a promising approach for simple, durable, and highly sensitive VFA measurement in cow rumen.
{"title":"Polymer-Coated Resonant Sensors for Monitoring Volatile Fatty Acids in the Cow Rumen","authors":"Fatin Bazilah Fauzi;Shintaro Noda;Michitaka Yamamoto;Jarred Fastier-Wooller;Yoshihiro Muneta;Shozo Arai;Toshihiro Itoh;Naoki Shiraishi","doi":"10.1109/LSENS.2025.3635713","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3635713","url":null,"abstract":"We developed polymer-coated resonant sensors to monitor volatile fatty acid (VFA) levels in the cow rumen. A 19-nm polybutadiene (PBD) film was applied to a quartz crystal microbalance (QCM), causing a frequency shift of approximately −400 Hz at 1 mM for all VFAs. As VFA concentrations increase, the frequency shift also increases in magnitude, highlighting the sensor's high sensitivity to VFAs. To test selectivity, a 171-nm-thick polydimethylsiloxane-coated QCM sensor was used. The frequency shifts observed were −178 Hz for acetic acid, −205 Hz for propionic acid, and −78 Hz for butyric acid, indicating different adsorption behaviors. Subsequently, we developed a VFA resonant sensor capsule for in vivo rumen monitoring. This capsule includes a PBD-coated QCM, a temperature/pressure sensor, an oscillator, a microcontroller, and a battery. In vivo tests showed a significant drop in resonant frequency within the cow rumen, demonstrating the detection of overall VFAs. The nano-thin polymer film coating on the resonant sensor demonstrates a promising approach for simple, durable, and highly sensitive VFA measurement in cow rumen.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"10 2","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-11-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145929351","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-18DOI: 10.1109/LSENS.2025.3634349
Xinyu Lu;Rui-qi Wang;Xiyou Sun;Lei Wang
This letter presents a multifunction electromagnetic probe for the simultaneous measurement of electric and magnetic fields and the enhancement of detection sensitivity. The multifunction composite probe consists of three main components: the sensing part, the transmission part, and the output part. The sensing section includes a pair of spiral loops acting as additional elements and a pair of U-shaped differential loops serving as driving elements. First, the U-shaped differential loops are used to achieve simultaneous testing of electric and magnetic fields. Second, a set of spiral loops as additional elements is introduced into the differential loops for sensitivity enhancement. Moreover, three connected vias are used to interconnect these differential loops with spiral loops: two different vias are utilized to connect the differential loops with the spiral loops, while a middle via is utilized to connect these spiral loops. To demonstrate the superiority of this design, the proposed electromagnetic probe is simulated and measured by high-frequency electromagnetic software and a near-field measurement system, respectively. Finally, experimental results show that the proposed multifunction electromagnetic probe has excellent advantages in multicomponent measurement (Hx and Ez) and high-sensitivity measurement.
{"title":"A Multifunction Electromagnetic Probe for the Simultaneous Measurement of Electric and Magnetic Fields and the Enhancement of Detection Sensitivity","authors":"Xinyu Lu;Rui-qi Wang;Xiyou Sun;Lei Wang","doi":"10.1109/LSENS.2025.3634349","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3634349","url":null,"abstract":"This letter presents a multifunction electromagnetic probe for the simultaneous measurement of electric and magnetic fields and the enhancement of detection sensitivity. The multifunction composite probe consists of three main components: the sensing part, the transmission part, and the output part. The sensing section includes a pair of spiral loops acting as additional elements and a pair of U-shaped differential loops serving as driving elements. First, the U-shaped differential loops are used to achieve simultaneous testing of electric and magnetic fields. Second, a set of spiral loops as additional elements is introduced into the differential loops for sensitivity enhancement. Moreover, three connected vias are used to interconnect these differential loops with spiral loops: two different vias are utilized to connect the differential loops with the spiral loops, while a middle via is utilized to connect these spiral loops. To demonstrate the superiority of this design, the proposed electromagnetic probe is simulated and measured by high-frequency electromagnetic software and a near-field measurement system, respectively. Finally, experimental results show that the proposed multifunction electromagnetic probe has excellent advantages in multicomponent measurement (<italic>H<sub>x</sub></i> and <italic>E<sub>z</sub></i>) and high-sensitivity measurement.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"9 12","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778220","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-17DOI: 10.1109/LSENS.2025.3633998
Federico G. De Geronimo;Nilton O. Renno
Silicon–graphene photodetectors have high gain, which varies with the environmental conditions. We developed a stabilization circuit that mitigates this problem by maintaining the graphene operating current nearly constant. We show that prototype devices with this circuit have a 2–4× smaller dark current noise power across temperatures ranging from −5 °C to 55 °C, and ∼4× lower noise power under modulated laser forcing with parasitic background white light. The device responsivity remains high under the changing environmental conditions, demonstrating that the proposed stabilization circuit significantly improves the performance of graphene–silicon photodetectors, enabling applications, such as optical tracking, alignment, and low-flux sensing.
{"title":"Low-Noise Stabilization Mechanism to Enhance the Performance of High-Gain Silicon–Graphene Photodetectors","authors":"Federico G. De Geronimo;Nilton O. Renno","doi":"10.1109/LSENS.2025.3633998","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3633998","url":null,"abstract":"Silicon–graphene photodetectors have high gain, which varies with the environmental conditions. We developed a stabilization circuit that mitigates this problem by maintaining the graphene operating current nearly constant. We show that prototype devices with this circuit have a 2–4× smaller dark current noise power across temperatures ranging from −5 °C to 55 °C, and ∼4× lower noise power under modulated laser forcing with parasitic background white light. The device responsivity remains high under the changing environmental conditions, demonstrating that the proposed stabilization circuit significantly improves the performance of graphene–silicon photodetectors, enabling applications, such as optical tracking, alignment, and low-flux sensing.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"9 12","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145674797","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-17DOI: 10.1109/LSENS.2025.3633315
Subham Koley;Sunil K Khijwania
This research aims to develop simple and novel optical fiber relative humidity (RH) sensor that employs intensity modulation via evanescent wave (EW) absorption. Proposed sensor exploits Al2O3/GO nanocomposite doped thin film of nanostructured silica as the sensing cladding on a centrally decladded plastic-clad silica (PCS) fiber. This configuration is used for the first time, to the best of the author's knowledge, for the development of optical fiber RH sensor. Comprehensive experimental investigations are carried out to establish response characteristics of the sensor. Proposed sensor demonstrates a significantly enhanced sensitivity of 0.0107 RH-1 while responding linearly over a dynamic range of 14%–86% RH. In addition, fast response/recovery time, excellent reversibility, repeatability, and reliability characteristics of the sensor make it suitable for real-field applications.
{"title":"Novel Al2O3/GO Nanocomposite-Based Highly Sensitive Optical Fiber Humidity Sensor","authors":"Subham Koley;Sunil K Khijwania","doi":"10.1109/LSENS.2025.3633315","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3633315","url":null,"abstract":"This research aims to develop simple and novel optical fiber relative humidity (RH) sensor that employs intensity modulation via evanescent wave (EW) absorption. Proposed sensor exploits Al<sub>2</sub>O<sub>3</sub>/GO nanocomposite doped thin film of nanostructured silica as the sensing cladding on a centrally decladded plastic-clad silica (PCS) fiber. This configuration is used for the first time, to the best of the author's knowledge, for the development of optical fiber RH sensor. Comprehensive experimental investigations are carried out to establish response characteristics of the sensor. Proposed sensor demonstrates a significantly enhanced sensitivity of 0.0107 RH<sup>-1</sup> while responding linearly over a dynamic range of 14%–86% RH. In addition, fast response/recovery time, excellent reversibility, repeatability, and reliability characteristics of the sensor make it suitable for real-field applications.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"10 1","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145729303","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-13DOI: 10.1109/LSENS.2025.3629702
Siyang Liu;Zijie Chen;Yiming Gao;Junrui Liang
People counting constitutes a crucial application of Internet of Things (IoT) technology. It offers valuable information for crowd management, security, and public health purposes. However, the majority of the current people counting sensors are powered either by batteries or by mains electricity. These power sources involve intricate installation procedures that frequently necessitate redecoration and arduous maintenance. This letter introduces a novel battery-free wireless floor tile sensor system for people counting. The floor tile terminal is composed of four quasi-static-toggling electromagnetic motion-powered switches. The foot traffic data transmitted are received by a gateway and subsequently forwarded to a cloud platform for analysis. The battery-free wireless floor tile is convenient to install. The entire system is capable of monitoring the number of people and their flow direction in real time. A prototype system is manufactured and installed at the entrance of the authors' laboratory for a field test. It achieves a 94.8% accuracy in walking directional identification and people counting. It is energy autonomy, low cost, and easy deployment. This study establishes a sustainable model for long-term indoor occupancy monitoring and crowd management. The design aligns with the current trend of eco-friendly, battery-free ambient IoT.
{"title":"Battery-Free Wireless Floor Tile for People Counting","authors":"Siyang Liu;Zijie Chen;Yiming Gao;Junrui Liang","doi":"10.1109/LSENS.2025.3629702","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3629702","url":null,"abstract":"People counting constitutes a crucial application of Internet of Things (IoT) technology. It offers valuable information for crowd management, security, and public health purposes. However, the majority of the current people counting sensors are powered either by batteries or by mains electricity. These power sources involve intricate installation procedures that frequently necessitate redecoration and arduous maintenance. This letter introduces a novel battery-free wireless floor tile sensor system for people counting. The floor tile terminal is composed of four quasi-static-toggling electromagnetic motion-powered switches. The foot traffic data transmitted are received by a gateway and subsequently forwarded to a cloud platform for analysis. The battery-free wireless floor tile is convenient to install. The entire system is capable of monitoring the number of people and their flow direction in real time. A prototype system is manufactured and installed at the entrance of the authors' laboratory for a field test. It achieves a 94.8% accuracy in walking directional identification and people counting. It is energy autonomy, low cost, and easy deployment. This study establishes a sustainable model for long-term indoor occupancy monitoring and crowd management. The design aligns with the current trend of eco-friendly, battery-free ambient IoT.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"9 12","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145560704","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-12DOI: 10.1109/LSENS.2025.3632112
Vinicius de Carvalho;Andre Lazzaretti;Marcia Muller;José Luís Fabris
This work presents the monitoring of torsion in a flexible cylindrical structure instrumented with a helically wound optical fiber. The sensing element consists of a standard fiber embedded in elastomer, forming a macrobend-based structure. Controlled angular displacements from $-90^{circ }$ to $90^{circ }$ were applied by twisting the structure. Distinct torsional states produced differentiable transmission spectra, with counterclockwise torsion increasing and clockwise torsion decreasing the mean transmittance across 475–750 nm. Single-wavelength fits showed wavelength-dependent behavior and limited predictive accuracy, highlighting the advantages of multivariate approaches that use full-spectrum information. Multivariate regression models were trained on spectral data reduced by principal component analysis for torsion prediction, with the elastic net achieving the best performance ($R^{2} = 0.99$). Residual analysis showed that 95% of prediction errors were below $3.5^{circ }$ for the 15-cm-long structure. These results confirm the feasibility of the proposed method for torsion sensing in soft robotic devices.
{"title":"Torsion Monitoring With a Helically Wound Macrobend Optical Fiber Sensor","authors":"Vinicius de Carvalho;Andre Lazzaretti;Marcia Muller;José Luís Fabris","doi":"10.1109/LSENS.2025.3632112","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3632112","url":null,"abstract":"This work presents the monitoring of torsion in a flexible cylindrical structure instrumented with a helically wound optical fiber. The sensing element consists of a standard fiber embedded in elastomer, forming a macrobend-based structure. Controlled angular displacements from <inline-formula><tex-math>$-90^{circ }$</tex-math></inline-formula> to <inline-formula><tex-math>$90^{circ }$</tex-math></inline-formula> were applied by twisting the structure. Distinct torsional states produced differentiable transmission spectra, with counterclockwise torsion increasing and clockwise torsion decreasing the mean transmittance across 475–750 nm. Single-wavelength fits showed wavelength-dependent behavior and limited predictive accuracy, highlighting the advantages of multivariate approaches that use full-spectrum information. Multivariate regression models were trained on spectral data reduced by principal component analysis for torsion prediction, with the elastic net achieving the best performance (<inline-formula><tex-math>$R^{2} = 0.99$</tex-math></inline-formula>). Residual analysis showed that 95% of prediction errors were below <inline-formula><tex-math>$3.5^{circ }$</tex-math></inline-formula> for the 15-cm-long structure. These results confirm the feasibility of the proposed method for torsion sensing in soft robotic devices.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"9 12","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-11-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145674710","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-07DOI: 10.1109/LSENS.2025.3630197
Ebrahim Nehary;Sreeraman Rajan
Phonocardiogram (PCG) can be used to detect cardiac conditions and support the initial diagnosis of cardiovascular disease, a critical health issue that requires early detection to allow timely treatment and potentially save lives. Classification of PCG signals as normal or abnormal is currently done using learning algorithms which require homogeneous training data. However, PCG datasets are often collected using stethoscopes with varying characteristics, from different individuals, and in diverse controlled or uncontrolled environments. This results in dataset heterogeneity, which poses a challenge for training effective deep learning models. This study explores the recently proposed Kolmogorov–Arnold Networks (KANs), which incorporate different trainable function families such as splines and wavelets for the classification of PCG and evaluate their robustness against data heterogeneity. KAN is compared with a traditional Multilayer Perceptron (MLP) on heterogeneous and homogeneous PCG datasets to determine the most suitable model for PCG classification. Experimental results show that KAN with wavelet-based functions outperforms KAN with spline functions and MLP on both datasets, achieving superior performance with parameters and computational costs comparable to those of MLP. In contrast, the spline-based KAN performs well on homogeneous data but poorly on heterogeneous data, incurring the highest computational cost and model complexity. KAN with wavelet functions outperforms MLP by over 10% in most cases and outperforms state-of-the art methods. In summary, KAN with wavelet functions demonstrate strong performance across dataset types and may be a promising candidate for fully connected layers in deep learning models, irrespective of whether the dataset is homogeneous or heterogeneous.
{"title":"Phonocardiogram Classification Model With Kolmogorov–Arnold Network for Training With Heterogeneous Dataset","authors":"Ebrahim Nehary;Sreeraman Rajan","doi":"10.1109/LSENS.2025.3630197","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3630197","url":null,"abstract":"Phonocardiogram (PCG) can be used to detect cardiac conditions and support the initial diagnosis of cardiovascular disease, a critical health issue that requires early detection to allow timely treatment and potentially save lives. Classification of PCG signals as normal or abnormal is currently done using learning algorithms which require homogeneous training data. However, PCG datasets are often collected using stethoscopes with varying characteristics, from different individuals, and in diverse controlled or uncontrolled environments. This results in dataset heterogeneity, which poses a challenge for training effective deep learning models. This study explores the recently proposed Kolmogorov–Arnold Networks (KANs), which incorporate different trainable function families such as splines and wavelets for the classification of PCG and evaluate their robustness against data heterogeneity. KAN is compared with a traditional Multilayer Perceptron (MLP) on heterogeneous and homogeneous PCG datasets to determine the most suitable model for PCG classification. Experimental results show that KAN with wavelet-based functions outperforms KAN with spline functions and MLP on both datasets, achieving superior performance with parameters and computational costs comparable to those of MLP. In contrast, the spline-based KAN performs well on homogeneous data but poorly on heterogeneous data, incurring the highest computational cost and model complexity. KAN with wavelet functions outperforms MLP by over 10% in most cases and outperforms state-of-the art methods. In summary, KAN with wavelet functions demonstrate strong performance across dataset types and may be a promising candidate for fully connected layers in deep learning models, irrespective of whether the dataset is homogeneous or heterogeneous.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"9 12","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145674715","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：