Decoding HD-EMG Signals for Myoelectric Control-How Small Can the Analysis Window Size be?

Rami N. Khushaba, Kianoush Nazarpour

Research output: Contribution to journalArticlepeer-review

Abstract / Description of output

Recently the use of high-density electromyogram (HD-EMG) signal acquisition setups has been promoted for myoelectric control and several databases have been made open access for scientific research. Despite this accelerated growth in the literature, coupled with industry interest, some fundamental research questions are unanswered. For instance, can HD-EMG signals be decoded at smaller window sizes and, if yes, what is the additional value of sophisticated motor unit decomposition methods. We used three open access databases, with 128 (8 and 65 hand movements) to 256 electrodes (34 hand movements) to explore the impact of varying the number of electrodes and segmentation windows size on EMG decoding accuracy. The analysis considered varying windows sizes from 32 ms to 256 ms and electrodes from 8 to 128/256. Simple time-domain and auto-regressive model parameters were extracted to train a linear discriminant analysis classifier to identify the intended hand motions. Our analysis indicates that with HD-EMG, even with windows sizes as small as 32 ms, one can achieve considerably high decoding accuracy, e.g. ≥90% , in our selected databases. As such, HD-EMG setups allow significant reductions in the analysis windows lengths up to 32 ms without compromising decoding performance.
Original languageEnglish
Pages (from-to)8569-8574
JournalIEEE Robotics and Automation Letters
Volume6
Issue number4
Early online date10 Sept 2021
DOIs
Publication statusPublished - 1 Oct 2021

Keywords / Materials (for Non-textual outputs)

  • Myoelectric control
  • high-density EMG
  • temporal-spatial interaction

Fingerprint

Dive into the research topics of 'Decoding HD-EMG Signals for Myoelectric Control-How Small Can the Analysis Window Size be?'. Together they form a unique fingerprint.

Cite this