Multiclass Detection and Tracking of Transient Motor Activation based on Decomposed Myoelectric Signals

Tutkimustuotos: Artikkeli kirjassa/konferenssijulkaisussavertaisarvioitu

Tutkijat

  • Martyna Stachaczyk
  • Seyed Farokh Atashzar
  • Sigrid Dupan
  • Ivan Vujaklija

  • Dario Farina

Organisaatiot

  • Imperial College London
  • Newcastle University

Kuvaus

Performance and efficacy of neuroprosthetic devices depend critically on the ability to detect the users motor intent with high temporal resolution. Delayed and incorrect responses significantly reduce usability, controllability and intuitiveness of prosthetic systems. Substantial efforts have been conducted to detect the steady-state phase of motor intention. However, detection, classification, and tracking of transient phases for one complete muscle contraction is still not possible. Clinically-established control systems rely mainly on surface electromyography (sEMG) signals in stationary, steady-state contractions, that have limited temporal resolution. Characterization of neural activities during different stages of a dynamic, transient contraction would allow for the development of a clinically-viable myoelectric system with a high temporal resolution that can significantly enhance the level of intuitiveness and usability of prosthetic devices. This could increase the response bandwidth and realize natural and dexterous control while avoiding exaggerated compensatory movements. For this purpose, in this paper, we explore the use of motor unit action potential trains (MUAPTs) for designing a movement intention detection technique. The goal is to classify and track the transient phases of muscle activation. Data collected from three subjects, during flexion tasks with four individual digits, is used to evaluate the algorithm. The performance is compared with that of the standard sEMG-based approach. Results showed a substantial advantage of the MUAPT-based phase detection algorithm over the conventional sEMG-based technique. It is confirmed that decoding, classification, and tracking of all stages of a dynamic, transient contraction is feasible using the proposed MUAPT-based approach, as a robust and efficient alternative for conventional sEMG-based algorithms.

Yksityiskohdat

AlkuperäiskieliEnglanti
Otsikko9th International IEEE EMBS Conference on Neural Engineering, NER 2019
TilaJulkaistu - 16 toukokuuta 2019
OKM-julkaisutyyppiA4 Artikkeli konferenssijulkaisuussa
TapahtumaInternational IEEE/EMBS Conference on Neural Engineering - San Francisco, Yhdysvallat
Kesto: 20 maaliskuuta 201923 maaliskuuta 2019
Konferenssinumero: 9

Julkaisusarja

NimiInternational IEEE/EMBS Conference on Neural Engineering
KustantajaIEEE
ISSN (painettu)1948-3546
ISSN (elektroninen)1948-3554

Conference

ConferenceInternational IEEE/EMBS Conference on Neural Engineering
LyhennettäNER
MaaYhdysvallat
KaupunkiSan Francisco
Ajanjakso20/03/201923/03/2019

ID: 34641038