A multimodal multistream multilevel fusion network for finger joint angle estimation with hybrid sEMG and FMG sensing

Alexandria Engineering Journal, Journal Year: 2024, Volume and Issue: 110, P. 9 - 23

Published: Oct. 5, 2024

Language: Английский

---

TinyML for Real-Time Embedded HD-EMG Hand Gesture Recognition with On-Device Fine-Tuning

Published: July 15, 2024

This work introduces a fully embedded wireless platform that incorporates the Coral Tensor Processing Unit (TPU) accelerator to leverage TinyML for real-time hand gesture recognition using high-density surface electromyography (HD-sEMG). With general inference time of 2.96 ms 64 channels sensor, TPU proved be well suited such tasks. Constructed from off-the-shelf components, offers cost-effective and self-sufficient alternative integrating artificial intelligence into prosthetic devices, eliminating dependency on expensive external hardware. The system allows intuitive calibration through user interface, facilitating fine-tuning model directly device or remotely via cloud-based server. On-device finetuning yields similar performance approach, improving accuracy by up 36.15% in intersession test cases. Extensive exploration 8-bit data quantization techniques demonstrates hardware compatibility can achieved without sacrificing performance. In best case, proposed scheme improve results 0.96% compared unquantized data. Overall, this paper establishes robust foundation advancing on-device HD-sEMG based recognition, paving way more accessible practical myoelectric solutions.

Language: Английский

---

Multitask Learning for Simultaneous Gesture and Force Level Recognition Toward Prosthetic Hand Interaction

IEEE Sensors Journal, Journal Year: 2024, Volume and Issue: 24(7), P. 11759 - 11769

Published: Feb. 9, 2024

In the field of intelligent prosthetics (IPs), establishment a natural interaction between prosthetic hands and amputees holds significant importance in restoring hand functionality, enhancing quality life, facilitating daily activities social engagement. Prior investigations on surface electromyographic (sEMG) signals-controlled IPs have predominantly concentrated gesture recognition, frequently neglecting equally dimension force level. This study proposes control strategy integrating multitask learning (MTL) model to achieve synchronized recognition gestures levels. The MTL model, incorporating shared convolutional blocks, self-attention, multihead attention layers, enhances for seamless user-device interaction. consistently showcases exceptional proficiency recognizing levels by conducting meticulous experimentation validating findings using datasets from diverse participants. Comparative assessments endorse superiority approach, particularly real-time testing scenarios. highlight potential this innovative myoelectric strategy, empowering users prompt, precise, intuitive responses, significantly augmenting their autonomy life.

Language: Английский

---

Shoulder Movement-Centered Measurement and Estimation Scheme for Underarm-Throwing Motions

Sensors, Journal Year: 2024, Volume and Issue: 24(10), P. 2972 - 2972

Published: May 7, 2024

Underarm throwing motions are crucial in various sports, including boccia. Unlike healthy players, people with profound weakness, spasticity, athetosis, or deformity the upper limbs may struggle find it difficult to control their hands hold release a ball using fingers at proper timing. To help them, our study aims understand underarm motions. We start by defining intention terms of launch angle ball, which goes hand-in-hand timing for releasing ball. Then, an appropriate part body is determined order estimate ball-throwing based on swinging motion. Furthermore, geometric relationship between movements and investigated involving multiple subjects. Based confirmed correlation, calibration-and-estimation model that considers individual differences proposed. The proposed consists calibration estimation modules. begin, as module performed, prediction states each subject updated online. module, estimated employing prediction. verify effectiveness model, extensive experiments were conducted seven In detail, two evaluation directions set: (1) how many balls need be thrown advance achieve sufficient accuracy; (2) whether can reach accuracy despite differences. From tests, 20 advance, could account estimation. Consequently, was when focusing shoulder human during throwing. near future, we expect expand means supporting disabled disabilities.

Language: Английский

---

3D-Printed Conductive Thermoplastic Electromyography Electrodes

Published: May 13, 2024

In the evolving landscape of assistive technologies, significant advancements are being made in functionality intelligent myoelectric prostheses, positioning them as a legitimate option for amputees and persons with congenital limb differences.Concurrently, 3D printing is transitioning from its traditional role prototyping tool to viable, cost-effective method manufacturing.Against this backdrop, it becomes feasible assess capabilities fabricating intricate components, such electrodes, which critical effective operation these prostheses.This study explores efficacy 3D-printed electrodes by producing evaluating three variants graphitedoped thermoplastic subsequently enhanced layer gold-plating.These innovative were benchmarked against five conventional electromyography (EMG) compare their performance characteristics.Testing ten participants revealed that two materials examined, PLA TPU, exhibited real potential applications.Notably, application goldplating thermoplastics significantly signal quality, achieving parity metal certain cases.This investigation underscores promising future doped medical applications.By enabling production combine conductive core an insulating exterior, technology paves way creation highly complex electrode designs.Moreover, ability rapidly prototype iterate designs through set revolutionize development process arrays, offering new avenues innovation not only prosthetic technology, but many other fields too.

Language: Английский

---