人机交互在康复任务中越来越受到关注，例如控制假肢或机器人手臂。利用表面肌电信号（sEMG）信号进行手势识别是最有前途的方法之一，因为sEMG信号采集是无创的并且与肌肉收缩直接相关。然而，这些信号的分析仍然存在许多挑战，因为相似的手势会导致相似的肌肉收缩。因此，得到的信号形状几乎相同，导致低分类精度。为了应对这一挑战，采用了复杂的神经网络，这需要大量的内存，消耗相对较高的能量，并限制了用于分类的设备的最大电池寿命。这项工作解决了这个问题，引入了生物成形剂。这一新的超小型注意力架构系列接近最先进的性能，同时减少了4.9倍的参数和操作数量。此外，通过引入新的受试者间预训练，我们将我们最好的Bioformer的准确性提高了3.39%，与最先进的准确性相匹配，而无需任何额外的推理成本。在并行超低功耗（PULP）微控制器单元（MCU）GreenWaves GAP 8上部署我们性能最好的Bioformer，我们实现了推理延迟和能量分别为2.72 ms和0.14 mJ，比之前最先进的神经网络低8.0倍，同时仅占用94.2 kB内存。
英文摘要：
Human-machine interaction is gaining traction in rehabilitation tasks, such as controlling prosthetic hands or robotic arms. Gesture recognition exploiting surface electromyographic (sEMG) signals is one of the most promising approaches, given that sEMG signal acquisition is non-invasive and is directly related to muscle contraction. However, the analysis of these signals still presents many challenges since similar gestures result in similar muscle contractions. Thus the resulting signal shapes are almost identical, leading to low classification accuracy. To tackle this challenge, complex neural networks are employed, which require large memory footprints, consume relatively high energy and limit the maximum battery life of devices used for classification. This work addresses this problem with the introduction of the Bioformers. This new family of ultra-small attention-based architectures approaches state-of-the-art performance while reducing the number of parameters and operations of 4.9X. Additionally, by introducing a new inter-subjects pre-training, we improve the accuracy of our best Bioformer by 3.39%, matching state-of-the-art accuracy without any additional inference cost. Deploying our best performing Bioformer on a Parallel, Ultra-Low Power (PULP) microcontroller unit (MCU), the GreenWaves GAP8, we achieve an inference latency and energy of 2.72 ms and 0.14 mJ, respectively, 8.0X lower than the previous state-of-the-art neural network, while occupying just 94.2 kB of memory.
TEMGNet: Deep Transformer-based Decoding of Upperlimb sEMG for Hand Gestures Recognition
Concordia University（康考迪亚大学）；Concordia Institute for Information System Engineering (CIISE)（康考迪亚大学信息系统工程研究所（CIISE））；Electrical and Computer Engineering (ECE), Concordia University（康考迪亚大学电气与计算机工程系（ECE））；Department of Bioengineering, Imperial College London（帝国理工学院生物工程系）；Electrical and Computer Engineering and Mechanical and Aerospace Engineering, New York University (NYU)（纽约大学（NYU）电气与计算机工程及机械与航空航天工程系）；New York University (NYU)（纽约大学（NYU））；Imperial College London（帝国理工学院）
Artificial Intelligence Thrust, HKUST (Guangzhou)（香港科技大学（广州）人工智能方向）；School of Computing and Information Technology, Great Bay University（湾区大学计算机与信息技术学院）；Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences（中国科学院深圳先进技术研究院）；Department of Computer Science and Engineering, HKUST（香港科技大学计算机科学与工程系）；Guangzhou HKUST Fok Ying Tung Research Institute（广州香港科技大学霍英东研究院）
A Deep Learning Sequential Decoder for Transient High-Density Electromyography in Hand Gesture Recognition Using Subject-Embedded Transfer Learning
University of California Los Angeles (UCLA)（加州大学洛杉矶分校（UCLA））；New York University (NYU)（纽约大学（NYU））；NYU WIRELESS（纽约大学无线研究中心）；NYU Center for Urban Science and Progress (CUSP)（纽约大学城市科学与进步中心（CUSP））
User Training with Error Augmentation for Electromyogram-based Gesture Classification
Northeastern University（东北大学）；Khoury College of Computer Sciences（科里计算机科学学院）；Meta Reality Labs Research（Meta现实实验室）；Institute for Experiential AI（体验式人工智能研究院）；Department of Physical Therapy, Movement, and Rehabilitation Sciences（物理治疗、运动与康复科学系）
Hand Gesture Recognition Using Temporal Convolutions and Attention Mechanism
Concordia University（康考迪亚大学）；Concordia Institute for Information System Engineering（康考迪亚信息系统工程研究院）；Imperial College London（帝国理工学院）；New York University（纽约大学）
Integrated Sensing and Communications for Real-time Avatar Control in XR over 5G
IDLab, University of Antwerp - imec（ID实验室，安特卫普大学 - imec）；IDLab, Ghent University - imec（ID实验室，根特大学 - imec）；Lighting Technology Lab, KU Leuven（照明技术实验室，鲁汶大学）
HYDRA-HGR: A Hybrid Transformer-based Architecture for Fusion of Macroscopic and Microscopic Neural Drive Information
Concordia University（康考迪亚大学）；Concordia Institute of Information Systems Engineering (CIISE)（康考迪亚信息系统工程研究院）；New York University (NYU)（纽约大学）；UNSW SYDNEY（悉尼新南威尔士大学）
Event-Driven Implementation of a Physical Reservoir Computing Framework for superficial EMG-based Gesture Recognition
James Watt School of Engineering, University of Glasgow（格拉斯哥大学詹姆斯·瓦特工程学院）；Institute of Neuroinformatics, University of Zurich and ETH Zurich（苏黎世大学神经信息研究所）；University of Glasgow（格拉斯哥大学）；School of Science and Engineering, University of Dundee（邓迪大学科学与工程学院）
