Get exclusive breakthroughs on sensors in IoT, energy, healthcare, and more, delivered straight to your inbox.
"IEEE Sensors Alert" is a pilot project of the IEEE Sensors Council. Started as one of its new initiatives, this weekly digest publishes teasers and condensed versions of our journal papers in layperson's language.
Articles Posted in the Month (December 2024)
Lightweight Inception Networks for the Recognition and Detection of Rice Plant Diseases
Published in: IEEE Sensors Journal (Volume: 22, Issue: 14, July 2022)
Summary Contributed by: Payal Savani
Rice, a Global staple, is often plagued by various diseases that impact crop production. The challenge of identifying these diseases exacerbates the issue. While deep learning is a powerful tool in image processing and computer vision, its application in plant disease recognition has been restricted. This paper introduces MobInc-Net, a lightweight Inception network that recognizes and detects rice plant diseases. It offers a practical solution that achieves high accuracy even in challenging conditions.
Self-Driven Photodetectors Based on Flexible Silicon Nanowires Array Surface-Passivated With Tin-Based Perovskites
Author: Yang Shengyi, Ge Zhenhua, Jiang Yurong, Wang Ying, Xin Haiyuan, Zhang Zhenheng, Zou Bingsuo
Published in: IEEE Sensors Journal (Volume: 24, Issue: 14, July 2024)
Summary Contributed by: Shengyi Yang (Author)
Silicon nanowire (Si-NW) photodetectors show great potential as efficient, self-driven, and compact devices in optoelectronic applications. However, their intrinsic surface defects reduce their responsivity and specific detectivity, limiting their performance. This paper introduces a novel self-driven photodetector based on flexible silicon nanowires array surface passivated with tin-based perovskites (FASnBr₃). The innovative design significantly enhances device performance and flexibility, making it a promising candidate for next-generation photodetectors.
Published in: IEEE Sensors Journal (Volume: 23, Issue: 3, February 2023)
Summary Contributed by: Saurabh Dubey
Anomalies between trains and platform doors threaten intercity railway safety. The paper proposes a method for anomaly detection using train predeparture key frame extraction and an Image-inpainting Anomaly Detection Network (IADN) based on image-inpainting autoencoder (AE) and local abnormal information enhancement and global-attentive reconstruction error (GARE). The tested results show effective and accurate anomaly detection, even outperforming state-of-the-art methods, ensuring safety with potential applications in security and locomotive industries.
Published in: IEEE Sensors Journal (Volume: 24, Issue: 15, August 2024)
Summary Contributed by: Nhien-An Le-Khac (Author)
Human activity recognition (HAR) using multiple sensors offers higher accuracy but raises privacy and convenience issues, while single sensors often lack detail and accuracy. The paper proposes Virtual Fusion with Contrastive Learning (VFCL), a novel framework for single-sensor-based activity recognition. Virtual fusion uses data from multiple sensors across different modalities for training but requires only one for predictions, while contrastive learning improves the accuracy and performance of each sensor independently.
Published in: IEEE Sensors Journal (Volume: 22, Issue: 13, July 2022)
Summary Contributed by: Saurabh Dubey
Early detection of breast cancer saves lives. The research presents a novel and adaptable breast cancer detection system integrating dual-polarized Ultra-Wideband (UWB) antennas on flexible Kapton polyimide, ensuring high precision. Eight UWB units surround the breast phantom and reconstruct 3D images using a delay-and-sum (DAS) algorithm to locate tumors with minimal clutter. Wearable and versatile, it can detect tumors with a 15 mm edge-to-edge distance, offering convenient health monitoring and self-diagnosis.
Design, Fabrication, and Validation of a Flexible Tactile Sensor for a Hand Prosthesis
Author: Kuo Chung-hsien, Nguyen Dai-Dong, Su Shun-Feng, Xie Wu-Qi
Published in: IEEE Sensors Journal (Volume: 24, Issue: 16, March 2024)
Summary Contributed by: Chung-Hsien Kuo (Author)
The design of a flexible tactile sensor using liquid metal (LM) and elastic fibers provides sensing capability and enhances the performance of a hand prosthesis. This study details the measurement principles of the LM-based force sensor, the design and fabrication process, and the sensor signal processing circuit. The proposed flexible tactile sensor offers reliable performance with high sensitivity in three axes, low error, and improved functionality in hand prostheses.
Published in: IEEE Sensors Journal (Volume: 22, Issue: 17, September 2022)
Summary Contributed by: Payal Savani
In our technology-driven world, devices require quick and efficient data processing. Edge computing enables rapid local decision-making, preserving bandwidth and privacy. Understanding platform intricacies is crucial for informed decision-making while navigating through vast data. The paper explores innovative approaches and experimental findings, providing insights into Deep Neural Networks (DNNs) architecture performance across diverse edge technologies. This aids in selecting optimal architectures based on performance metrics for specific applications.
Published in: IEEE Sensors Journal (Volume: 24, Issue: 13, July 2024)
Summary Contributed by: Arantxa Uranga (Author)
Hydrophones are devices that convert underwater acoustic pressure into electrical signals. The paper proposes a hydrophone designed using Aluminum Scandium Nitride (AlScN) piezoelectric micromachined ultrasonic transducers (PMUTs) integrated monolithically on CMOS (Complementary Metal-Oxide-Semiconductor). This single-chip AIScN PMUTs with COMS (PMUTs-on-CMOS) hydrophone offers compactness, high sensitivity, and energy efficiency for underwater acoustic sensing. It supports high-performance underwater detection and has promising applications in underwater communications, sonar, and environmental monitoring systems.
A Retail Object Classification Method Using Multiple Cameras for Vision-Based Unmanned Kiosks
Author: Ji-Ye Jeon, Shin-Woo Kang, Hyuk-Jae Lee, Jin-Sung Kim
Published in: IEEE Sensors Journal (Volume: 22, Issue: 22, November 2022)
Summary Contributed by: Saurabh Dubey
Sensor technology, wireless communications, and Internet of Things (IoT) use in unmanned self-checkout systems has elevated the retail experience. This paper presents a vision-based RGB kiosk using a sophisticated combination of multiple cameras and a cutting-edge Convolutional Neural Network (CNN) framework, yielding a 33.67% improvement over conventional methods. It solves inter-classification challenges and intra-class variations in products with 142k+ real-world dataset points, with promising applications in retail and lifestyle sectors.
TSSTDet: Transformation-Based 3-D Object Detection via a Spatial Shape Transformer
Author: Yoo Myungsik, Bui Cuong Duy, Hoang Hiep Anh
Published in: IEEE Sensors Journal (Volume: 24, Issue: 5, March 2024)
Summary Contributed by: Myungsik Yoo (Author)
Accurate 3D object detection is essential for the safe navigation of autonomous vehicles. The novel transformation-based 3-D object detection via a spatial shape transformer (TSSTDet) overcomes the challenges of incomplete and varying orientations of the obstructions. Its key features include a rotational transformation convolutional backbone (RTConv) for orientation-invariant detection and a voxel-point shape transformer for reconstructing missing parts, thus improving obstacle detection and avoidance and enhancing safe navigation in autonomous driving.
Maintaining indoor air quality (IAQ) is crucial for health and wellness. Accurate data analysis and contextual anomaly detection are essential for IAQ monitoring. The paper introduces a hybrid deep-learning model, combining long short-term memory (LSTM) with autoencoder (AE). LSTM learns typical carbon dioxide (CO2) time sequence patterns, while AE computes optimal reconstruction errors and detects anomalies. Achieving 99.50% accuracy in real-world testing, the model shows promise for enhancing IAQ monitoring.
Hand gesture recognition has become an integral part of Human-Computer Interactions. The paper introduces a methodology using a video-based dataset and convolutional neural network (CNN) model. It utilizes an RGB-Depth camera to create a dataset of six distinct hand gestures. A lightweight CNN model is then developed to detect and classify hand movements. The experimental results highlight its accuracy and efficiency, facilitating its practical use in scenarios demanding precise gesture recognition.
A non-profit organization, IEEE is the world's largest technical professional organization dedicated to advancing technology for the benefit of humanity.
Copyright 2023 IEEE – All rights reserved. Use of this website signifies your agreement to the IEEE Terms and Conditions
This site is also available on your smartphone.
