Review
Engineering, Electrical & Electronic
Lian Yue, Lu Zongxing, Dong Hui, Jia Chao, Liu Ziqiang, Liu Zhoujie
Summary: Researchers are exploring ways to enable machines to interpret human body language for more intelligent and efficient human-machine interaction. Foot gesture recognition has emerged as a popular technology with its simplicity, speed, and accuracy in capturing various gestures and conveying body information.
IEEE SENSORS JOURNAL
(2023)
Article
Chemistry, Multidisciplinary
Han Fang, Lei Wang, Zhongzheng Fu, Liang Xu, Wei Guo, Jian Huang, Zhong Lin Wang, Hao Wu
Summary: Recent advances in flexible wearable devices have led to the development of human-machine interfaces that are of great value to emerging cybernetics, robotics, and Metaverse systems. However, the effectiveness of existing approaches is limited by the quality of sensor data and computationally expensive classification models. In this study, a novel gesture recognition system with triboelectric smart wristbands and an adaptive accelerated learning (AAL) model is proposed. The system demonstrates highly sensitive and high-quality sensing capabilities, as well as achieving a 97.56% identification accuracy with reduced computational costs.
Article
Neurosciences
Mo Han, Mehrshad Zandigohar, Sezen Yagmur Gunay, Gunar Schirner, Deniz Erdogmus
Summary: This study proposes a framework for classifying dynamic EMG signals into gestures and examines the impact of different movement phases using an unsupervised segmentation and labeling method. Data from large gesture vocabularies were collected for encoding transitions between grasp intents based on natural sequences of human grasp movements. A classifier was constructed for identifying gesture labels based on dynamic EMG signals without the need for supervised annotation of kinematic movements. Various training strategies using EMG data from different movement phases were evaluated in real-time performance transitions.
FRONTIERS IN NEUROSCIENCE
(2022)
Article
Automation & Control Systems
Fady S. Botros, Angkoon Phinyomark, Erik J. Scheme
Summary: This study investigates the feasibility of hand gesture recognition using wrist EMG signals, finding that wrist signals have higher signal quality for gestures involving fine finger movements, while maintaining comparable quality for wrist gestures. The results suggest the potential of using wrist EMG signals in hand gesture recognition and the importance of incorporating knowledge from the prosthetics field into the design of EMG pattern recognition systems.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2022)
Article
Computer Science, Information Systems
Jubayer Al Mahmud, Bandhan Chandra Das, Jungpil Shin, Khan Md Hasib, Rifat Sadik, M. F. Mridha
Summary: The research proposes a 3D gesture recognition and adaption system based on Kinect for human-robot interaction, comprising four modules: pointing gesture recognition, 3D dynamic gesture recognition, gesture adaptation, and robot navigation. The system achieves high accuracy in dynamic gesture recognition and is capable of recognizing pointing gestures in real-time.
Article
Chemistry, Analytical
Jona Grosse Sundrup, Katja Mombaur
Summary: We explored the distribution of muscle signatures of human hand gestures using Dynamic Time Warping. A k-Nearest-Neighbors classifier was used with Dynamic Time Warping as the distance estimate to evaluate the classification performance. Additionally, we developed a method to assess the separability of a set of gestures by investigating the distribution of recorded samples. Furthermore, we proposed and evaluated two approaches with reduced computational cost for real-time usage, considering their effectiveness and underlying mechanics. We also examined the impact of different parameters on practical usability and background rejection to fine-tune the classification procedure.
Article
Robotics
Marc Peral, Alberto Sanfeliu, Anais Garrell
Summary: In this paper, an efficient and reliable deep-learning approach for hand gesture recognition in robot communication is presented. The approach utilizes visual information to extract hand landmarks and predict gestures, achieving high accuracy and real-time performance.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2022)
Review
Chemistry, Multidisciplinary
Zhihan Lv, Fabio Poiesi, Qi Dong, Jaime Lloret, Houbing Song
Summary: Gesture recognition and speech recognition, as important input methods in Human-Computer Interaction (HCI), have achieved breakthrough research progress with the rapid development of deep learning and artificial intelligence. This study analyzes the current situation of HCI intelligent systems and summarizes the implementation of gesture interaction and voice interaction in HCI. The combination of intelligent HCI and deep learning is deeply applied in gesture recognition, speech recognition, emotion recognition, and intelligent robot direction, resulting in high recognition accuracy and better robustness in Human-Machine Interfaces (HMIs) with voice support.
APPLIED SCIENCES-BASEL
(2022)
Article
Computer Science, Information Systems
Sathishkumar Subburaj, Chih-Ho Yeh, Brijesh Patel, Tsung-Han Huang, Wei-Song Hung, Ching-Yuan Chang, Yu-Wei Wu, Po Ting Lin
Summary: In this study, human gestures are classified using data from a curved piezoelectric sensor. Machine learning algorithms enabled with K-mer are developed and optimized to achieve the best results in human gesture recognition. Support vector machine, random forest, and k-nearest neighbor algorithms are analyzed and performed with K-mer. The proposed model achieves promising results in accuracy and has relatively fast execution time.
Article
Computer Science, Artificial Intelligence
Tugba Zeybek, Ufuk Sakarya
Summary: Human-machine interaction (HMI) is an important research area that has gained significance with the development of sensor technologies. Gesture recognition in unmanned aerial vehicles (UAVs) has become a considerable study in recent years. This paper aims to create a simple human-machine interface design for UAVs. The gesture recognition problem can be modeled as a pattern recognition problem in a multi-sensor system.
JOURNAL OF INTELLIGENT & ROBOTIC SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Valeria Villani, Cristian Secchi, Marco Lippi, Lorenzo Sabattini
Summary: Recent advances in robotics have made it possible for robots to assist and work together with humans. To promote their use and diffusion, intuitive and user-friendly interaction means should be adopted. Gestures have become an established way to interact with robots, and this article focuses on the problem of gesture recognition in human-robot interaction (HRI). The authors propose a pipeline framework that takes into account the specific constraints of HRI and evaluate its performance using standard machine learning algorithms.
IEEE TRANSACTIONS ON HUMAN-MACHINE SYSTEMS
(2023)
Article
Chemistry, Multidisciplinary
Puchuan Tan, Xi Han, Yang Zou, Xuecheng Qu, Jiangtao Xue, Tong Li, Yiqian Wang, Ruizeng Luo, Xi Cui, Yuan Xi, Le Wu, Bo Xue, Dan Luo, Yubo Fan, Xun Chen, Zhou Li, Zhong Lin Wang
Summary: In this study, a gesture recognition wristband that allows for full keyboard and multicommand input is developed. The wristband is convenient to wear, low in cost, and does not interfere with other hand movements. With the use of physiological anatomy, active sensor, and machine learning technology, the wristband achieves a high accuracy rate in recognizing 26 letters.
ADVANCED MATERIALS
(2022)
Article
Chemistry, Analytical
Yanru Bai, Xiaoqing Li, Chengcai Zheng, Rui Guo, Xisheng Li
Summary: Gesture recognition is crucial for intelligent control and human-computer interaction. A new type of stretchable EMG electrode based on liquid metal nanoparticles has been developed, which conforms well to the mechanical properties of human skin. This electrode has the advantages of low cost, great stretch ability, and high accuracy in collecting EMG signals generated by hand movements.
Article
Chemistry, Multidisciplinary
Marc Kurz, Robert Gstoettner, Erik Sonnleitner
Summary: The miniaturization of electronic components has led to the development of smart rings containing motion sensors, enabling natural interaction through hand gesture recognition. Smart rings are highly mobile and communicate wirelessly, potentially serving as touchless user interfaces for various applications. Through training with multiple machine learning models, the system achieves high accuracy scores for gesture recognition tasks.
APPLIED SCIENCES-BASEL
(2021)
Review
Automation & Control Systems
Dezhen Xiong, Daohui Zhang, Xingang Zhao, Yiwen Zhao
Summary: This paper examines the role of deep learning in decoding EMG signals for human-machine interaction applications. It reviews recent advancements in network structures, processing schemes, and tasks like movement classification and joint angle prediction. The discussion also includes new challenges, such as multimodal sensing and robustness towards disturbances, and presents potential future research directions.
IEEE-CAA JOURNAL OF AUTOMATICA SINICA
(2021)
Article
Polymer Science
Luca Michele Martulli, Riccardo Sala, Gennaro Rollo, Milutin Kostovic, Marino Lavorgna, Andrea Sorrentino, Emanuele Gruppioni, Andrea Bernasconi
Summary: This study investigates the potential of using three-dimensional printed polymers as an alternative to composite laminate materials for prosthetic feet. A numerical approach is proposed and validated, and a new component is designed based on a commercial prosthesis using additive technology. The results demonstrate that additive manufacturing can be a low-cost alternative for producing affordable prosthetic feet.
Article
Computer Science, Hardware & Architecture
Enrico Tabanelli, Giuseppe Tagliavini, Luca Benini
Summary: Machine Learning (ML) functions are widely used in latency-sensitive IoT applications, leading to an increasing adoption of Parallel Ultra-low-power (PULP) IoT processors for near-sensor processing. This article focuses on enabling efficient parallel execution of Non-neural ML algorithms on two RISCV-based PULP platforms, GAP8 and PULP-OPEN. The optimized parallel algorithms, with target-optimized emulation libraries and native FPU support, achieve significant runtime improvement and energy reduction compared to standard emulation libraries, and outperform the ARM Cortex-M4 microcontroller in terms of speed.
ACM TRANSACTIONS ON EMBEDDED COMPUTING SYSTEMS
(2023)
Article
Engineering, Electrical & Electronic
Gianmarco Ottavi, Angelo Garofalo, Giuseppe Tagliavini, Francesco Conti, Alfio Di Mauro, Luca Benini, Davide Rossi
Summary: Dustin is a fully programmable compute cluster with 16 RISC-V cores capable of 2-to-32-bit arithmetic and mixed-precision calculations. It introduces a Vector Lockstep Execution Mode (VLEM) to reduce power consumption in highly data-parallel kernels, achieving a 38% power reduction. Implemented in 65nm CMOS technology, the cluster achieves a peak performance of 58 GOPS and a peak efficiency of 1.15 TOPS/W.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I-REGULAR PAPERS
(2023)
Article
Computer Science, Hardware & Architecture
Nils Wistoff, Moritz Schneider, Frank K. Gurkaynak, Gernot Heiser, Luca Benini
Summary: Microarchitectural timing channels allow unauthorized information flow across security boundaries, violating fundamental security assumptions. This study introduces the temporal fence instruction fence.t using the open and extensible RISC-V instruction set architecture to prevent information leakage via timing channels. The most effective implementation involves erasing all non-architectural core components, supported by the ISA, with a low implementation effort, minimal performance overhead of less than 1%, and negligible hardware costs.
IEEE TRANSACTIONS ON COMPUTERS
(2023)
Article
Chemistry, Analytical
Andrea Amidei, Susanna Spinsante, Grazia Iadarola, Simone Benatti, Federico Tramarin, Paolo Pavan, Luigi Rovati
Summary: The majority of car accidents worldwide are caused by drowsy drivers. Therefore, it is important to be able to detect drowsiness in drivers to prevent serious accidents. This study focuses on using a wrist device and analyzing the skin conductance signal to detect drowsiness. The results show that it is possible to identify drowsiness accurately using only signals from the wrist, encouraging further research for a real-time warning system.
Article
Materials Science, Biomaterials
Andrea Giovanni Cutti, Federico Morosato, Emanuele Gruppioni, Gregorio Teti, Lorenzo De Michieli, Cosimo Gentile
Summary: Conventional myoelectric prosthetic hands only offer a basic tri-digital pinch. Transradial amputees need to compensate for this lack of function with altered kinematics at the shoulder and trunk that might expose them to an increased risk of musculoskeletal injuries. A poly-articulated prosthetic hand may reduce the physical compensatory movements and close the gap between the sound and the prosthetic side.
Article
Engineering, Electrical & Electronic
Alessandro Ottaviano, Thomas Benz, Paul Scheffler, Luca Benini
Summary: Power and cost constraints in the IoT extreme-edge and TinyML domains have led to a trend towards heterogeneous architectures. This brief presents Cheshire, a lightweight and modular 64-bit Linux-capable host platform designed for the seamless integration of domain-specific accelerators. Cheshire features a unique low-pin-count DRAM interface, a configurable cache, and a DMA engine for efficient data movement.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II-EXPRESS BRIEFS
(2023)
Article
Engineering, Electrical & Electronic
Matteo Perotti, Matheus Cavalcante, Alessandro Ottaviano, Jiantao Liu, Luca Benini
Summary: The nature and heterogeneity of modern workloads present a dilemma for hardware designers to choose between general-purpose processors and highly-tailored accelerators. The vector processor architecture, combining the flexibility of a processor with the performance and efficiency of an accelerator, is a promising solution. This article introduces the Yun SoC, the first implementation of an open-source RISC-V-based vector processor, which achieves impressive performance and energy efficiency.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II-EXPRESS BRIEFS
(2023)
Article
Computer Science, Hardware & Architecture
Bruno Sa, Luca Valente, Jose Martins, Davide Rossi, Luca Benini, Sandro Pinto
Summary: This article describes the author's work on hardware virtualization support in the RISC-V CVA6 core, including architecture, microarchitecture, and design space exploration. They design a set of microarchitectural enhancements, such as GTLB and L2 TLB, to alleviate virtualization performance overhead. Based on design space exploration and post-layout simulations, they select an optimal hardware configuration with a 16% performance speedup compared to nonoptimized design.
IEEE TRANSACTIONS ON VERY LARGE SCALE INTEGRATION (VLSI) SYSTEMS
(2023)
Article
Materials Science, Multidisciplinary
Kirill Keller, Christoph Leitner, Christian Baumgartner, Luca Benini, Francesco Greco
Summary: We present the fabrication of a fully printed, lead-free, polymer piezoelectric transducer and demonstrate the characterization of its structural, dielectric, and ferroelectric properties at different processing stages. The performance of the transducer made with poly(vinylidene fluoride-trifluoroethylene) is evaluated through resonance frequency analyses, acoustic power measurements, and pulse-echo experiments. Importantly, we demonstrate an optimal performance in the medical ultrasound range (<15 MHz) with acoustic power >1 W cm(-2) for the first time in a fully printed transducer, which shows promise for applications in epidermal and wearable electronics. Overall, these findings lay a strong foundation for future research in the field of flexible ultrasound transducers.
ADVANCED MATERIALS TECHNOLOGIES
(2023)
Article
Engineering, Electrical & Electronic
Amirhossein Moallemi, Riccardo Gaspari, Federica Zonzini, Luca De Marchi, Davide Brunelli, Luca Benini
Summary: Structural health monitoring systems use low-cost, low-energy sensors and processing units to monitor facilities, but dense sensor deployments generate large volumes of data that require high bandwidth and consume substantial power. This study introduces a parallelized data reduction method for vibration analysis, leveraging the capabilities of the GAP9 MCU to reduce execution time by approximately 60x and power consumption to just 48.3mW while preserving spectral accuracy.
IEEE SENSORS LETTERS
(2023)
Article
Computer Science, Information Systems
Tommaso Polonelli, Amirhossein Moallemi, Weikang Kong, Hanna Muller, Julien Deparday, Michele Magno, Luca Benini
Summary: This paper presents a low-power, self-sustainable, and time-synchronised system for aerodynamic and acoustic measurements on operating wind turbines. The system can acquire precise data with a time synchronisation error below 100 mu s and sustain a data rate of 600 kbps over 400 m with up to 5 sensor nodes.
Article
Polymer Science
Gennaro Rollo, Alfredo Ronca, Pierfrancesco Cerruti, Hesheng Xia, Emanuele Gruppioni, Marino Lavorgna
Summary: This study presents a novel piezoresistive sensor made by 3D printing with ultralight nature, high compressibility, robustness, and excellent electromechanical properties. The optimization of the sensor's performance is achieved by studying the variables such as thickness, diameter, and porosity of 3D printed disks. The resulting sensor shows optimal reproducibility, sensitivity, and linearity of the electrical signal.
Article
Computer Science, Theory & Methods
Paul Scheffler, Florian Zaruba, Fabian Schuiki, Torsten Hoefler, Luca Benini
Summary: Sparse linear algebra is important in many application domains, but it is challenging to handle efficiently. In this research, an existing memory-streaming RISC-V ISA extension was enhanced to accelerate the operations of sparse tensors, achieving higher speedups and efficiency.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
(2023)
Article
Robotics
Hanna Muller, Vlad Niculescu, Tommaso Polonelli, Michele Magno, Luca Benini
Summary: Nano size drones have potential for exploring unknown and complex environments, but their limitations in computation and sensing make full autonomy challenging. This article presents a lightweight obstacle avoidance system using a novel millimeter ToF sensor and a generalized control policy, achieving reliable performance in indoor environments.
IEEE TRANSACTIONS ON ROBOTICS
(2023)