FEASIBILITY OF SPATIO-TEMPORAL LINEAR FEATURE LEARNING FOR MYOELECTRIC CONTROL: A SMALL WINDOW SIZE APPROACH
DOI:
https://doi.org/10.57922/mec.2512Abstract
Numerous research papers have delved into spatio-temporal analysis for myoelectric control, yielding meaningful outcomes, often employing window sizes ranging from 100 to 300 milliseconds. However, the industry is interested in achieving robust performance within smaller window sizes, more applicable to real-world scenarios. This study introduces a novel approach, Spatio-Temporal Linear Feature Learning (STLFL), with a robust trade-off between high performance and compact window size. Our investigation primarily focused on five classes within the state-of-the-art DB5 dataset—rest, abduction of all fingers, pointing index, power sphere grasp, and prismatic pinch grasp. Comparative analyses with two established methods, namely support vector machine (SVM) and convolutional neural network (CNN), revealed that STLFL consistently outperformed, achieving an impressive average accuracy of 84.6±3.9% across 10 subjects within an 80-millisecond window in a 16-channel electromyography signal. These results highlight the efficiency of STLFL in achieving myoelectric control within a limited timeframe, demonstrating promising outcomes for multiclass applications in both future contexts and real-world scenarios.
Downloads
Published
2024-08-15
How to Cite
[1]
S. N. Aghili and K. Nazarpour, “FEASIBILITY OF SPATIO-TEMPORAL LINEAR FEATURE LEARNING FOR MYOELECTRIC CONTROL: A SMALL WINDOW SIZE APPROACH”, MEC Symposium, Aug. 2024.
Conference Proceedings Volume
Section
Myoelectric Control Algorithms
