The article discusses the development of a novel deep-learning model to enhance the function of high-density surface electromyography (HD-sEMG) modules utilized in neurorobotic systems. These modules are critical for controlling bionic limbs and prosthetic systems. However, they can be affected by issues such as electrode-skin disconnections and sensor dropout. The new model, named 3D Dilated Efficient CapsNet, is trained to learn channel dropout variations and thus increase robustness to channel dropout. It has shown high performance in a sensor dropout reliability study, demonstrating its potential for improving the reliability and scalability of neurorobotic systems.
Publication date: 21 Sep 2023
Project Page: Not provided
Paper: https://arxiv.org/pdf/2309.11086
