Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Myoelectric Control for Hand Prostheses

Sebelius, Fredrik LU orcid (2004)
Abstract
An investigation of improvements of myoelectric prostheses has been undertaken. The primary aims of this thesis were (1) to generate an accurate prediction of as many hand movement as possible, (2) to produce a training setup for subjects allowing intuitive and instant control over multiple movements, and (3) to reduce the training cycle for the control system to a maximum of a couple of minutes to enable optimizations, e.g., electrode placement. A median of six movements has been predicted with a 100% accuracy. At the initial predictions, a new set-up for training amputees using a data glove has been proposed, and training of less than 30 seconds of off-line learning, as well as direct online learning, has been conducted. Thus, the... (More)
An investigation of improvements of myoelectric prostheses has been undertaken. The primary aims of this thesis were (1) to generate an accurate prediction of as many hand movement as possible, (2) to produce a training setup for subjects allowing intuitive and instant control over multiple movements, and (3) to reduce the training cycle for the control system to a maximum of a couple of minutes to enable optimizations, e.g., electrode placement. A median of six movements has been predicted with a 100% accuracy. At the initial predictions, a new set-up for training amputees using a data glove has been proposed, and training of less than 30 seconds of off-line learning, as well as direct online learning, has been conducted. Thus, the initial goals were fulfilled. Further, an online learning system has proved to further increase the accuracy and the number of movements performed while the response time for prediction decreased to 50–100 ms. (Less)
Abstract (Swedish)
Popular Abstract in Swedish

Denna avhandling syftar till att förbättra styrningen av myoelektriska handproteser. De primära målen har varit (1) att generera korrekt kontroll för så många handrörelser som möjligt, (2) att förbättra träningsförfarandet så att operatören på ett intuitivt sätt snabbt kan bemästra ett flertal rörelser och (3) att minska tiden som behövs för att matematiskt justera in systemet till maximalt ett par minuter, så att man enkelt kan utföra olika typer av optimeringar, såsom omplacering av elektroder. En initial median noggrannhet på sex rörelser utav tio utförda med hundra procents korrekthet har hittills uppnåtts. En ny träningsuppställning för amputerade, där en datahandske (en handske med... (More)
Popular Abstract in Swedish

Denna avhandling syftar till att förbättra styrningen av myoelektriska handproteser. De primära målen har varit (1) att generera korrekt kontroll för så många handrörelser som möjligt, (2) att förbättra träningsförfarandet så att operatören på ett intuitivt sätt snabbt kan bemästra ett flertal rörelser och (3) att minska tiden som behövs för att matematiskt justera in systemet till maximalt ett par minuter, så att man enkelt kan utföra olika typer av optimeringar, såsom omplacering av elektroder. En initial median noggrannhet på sex rörelser utav tio utförda med hundra procents korrekthet har hittills uppnåtts. En ny träningsuppställning för amputerade, där en datahandske (en handske med vinkelmätare för lederna) monteras på den kontralaterala handen, har föreslagits. Data- handsken möjliggör snabb insamling av ledvinklar i handen simultant med elektromyografi av motsvarande muskelaktivitet. Vidare uppvisar de föreslagna igenkänningsalgoritmerna en inställningstid under 30 sekunder för förinspelad data på tre minuter (tio basrörelser) på en högst ordinär dator. Algoritmerna lämpar sig utmärkt för en realtidsapplikation, där justeringarna av systemet sker fortlöpande, då de ej behöver någon extra inställningstid. I denna utföringsform ökade noggrannheten och antal rätt klassificerade rörelser till mer än åtta utav tio standardrörelser, medan tidsfördröjningen mellan ”facit” och den predicerade rörelsen minskade till 50–100 ms (Less)
Please use this url to cite or link to this publication:
author
supervisor
opponent
  • Docent Wessberg, Johan, Inst. för fysiologi och farmakologi, Göteborgs universitet, Box 432, 405 30 Göteborg
organization
publishing date
type
Thesis
publication status
published
subject
keywords
Mät- och instrumenteringsteknik, Care and help to handicapped, Handikappade, vård och rehabilitering, Instrumentation technology, Virtual, Recognition, Real time, On-line learning, Myoelectric, Hand prosthesis, EMG, ANN, Artificial hands
pages
155 pages
publisher
Department of Electrical Measurements, Lund University
defense location
Room E:1406, E-building, Lund Institute of Technology
defense date
2004-04-02 10:15:00
external identifiers
  • other:ISRN:LUTEDX/TEEM--1078--SE
language
English
LU publication?
yes
additional info
Article: I Tele-Operation of Multi-Finger Virtual/Robot Hand Using Self-Organizing Tree Structured Network on EMGSebelius F, Eriksson L, Balkenius C, Laurell TProceedings of First International ICSC Conference on Neuro-Fuzzy Technologies NF'2002, 16–19 January, Havana, Cuba, 2002. Article: II Classification of Motor Commands Using a Modified Self-Organizing Feature MapSebelius F, Eriksson L, Holmberg H, Levinsson A, Lundborg G,Danielsen N, Schouenborg J, Balkenius C, Laurell T, Montelius LSubmitted (2004) to Medical Engineering & Physics. Article: III Motor Control of an Artificial Hand by Tree-Structured Network: A New Concept Based on the Combined Use of ANNs and a Data GloveSebelius F, Eriksson L, Balkenius C, Laurell TSubmitted (2003) to Journal of Medical Engineering & Technology. Article: IV Refined Motor Control of Artificial Hands: Preliminary Results from Six PatientsSebelius FCP, Rosén BN, Lundborg GNSubmitted (2003) to The Journal of Hand Surgery. Article: V Real-time Control of a Virtual HandSebelius F, Axelsson M, Danielsen N, Schouenborg J, Laurell TSubmitted (2004) to Technology and Disability.
id
d0b39306-6c5c-402e-9a09-1a812b4b324b (old id 466812)
date added to LUP
2016-04-04 11:04:50
date last changed
2024-03-12 12:41:58
@phdthesis{d0b39306-6c5c-402e-9a09-1a812b4b324b,
  abstract     = {{An investigation of improvements of myoelectric prostheses has been undertaken. The primary aims of this thesis were (1) to generate an accurate prediction of as many hand movement as possible, (2) to produce a training setup for subjects allowing intuitive and instant control over multiple movements, and (3) to reduce the training cycle for the control system to a maximum of a couple of minutes to enable optimizations, e.g., electrode placement. A median of six movements has been predicted with a 100% accuracy. At the initial predictions, a new set-up for training amputees using a data glove has been proposed, and training of less than 30 seconds of off-line learning, as well as direct online learning, has been conducted. Thus, the initial goals were fulfilled. Further, an online learning system has proved to further increase the accuracy and the number of movements performed while the response time for prediction decreased to 50–100 ms.}},
  author       = {{Sebelius, Fredrik}},
  keywords     = {{Mät- och instrumenteringsteknik; Care and help to handicapped; Handikappade; vård och rehabilitering; Instrumentation technology; Virtual; Recognition; Real time; On-line learning; Myoelectric; Hand prosthesis; EMG; ANN; Artificial hands}},
  language     = {{eng}},
  publisher    = {{Department of Electrical Measurements, Lund University}},
  school       = {{Lund University}},
  title        = {{Myoelectric Control for Hand Prostheses}},
  year         = {{2004}},
}