ETD system

Electronic theses and dissertations repository

 

Tesi etd-11122019-105752


Thesis type
Tesi di laurea magistrale
Author
CAPUCCIO, ANTONINO
URN
etd-11122019-105752
Title
Reducing the Effect of applied Loads and Limb Position on Myoelectric Hand Prosthesis Control using Accelerometers
Struttura
INGEGNERIA DELL'INFORMAZIONE
Corso di studi
BIONICS ENGINEERING
Supervisors
relatore Prof. Cipriani, Christian
correlatore Dott. Mannini, Andrea
controrelatore Prof. Tognetti, Alessandro
Parole chiave
  • Limb position effect
  • Inertial measurement units
  • Electromyography
  • Pattern-recognition
  • Hand prosthesis
  • Weight variation
Data inizio appello
06/12/2019;
Consultabilità
Secretata d'ufficio
Data di rilascio
06/12/2089
Riassunto analitico
Aiming at dexterous and reliable solutions to increase the quality of life of amputees, upper limb prosthetic devices and control methods are currently gaining interest in research. Among the different types of control approaches, myoelectric pattern-recognition based prosthesis have received a lot of attention in previous years, allowing the automatic recognition of complex movement intentions. However, the high rates of prosthesis abandonment that are observed in users that adopt this control approach still concern the scientific community. Several factors were identified in literature as cause for the loss in robustness and increase in EMG pattern classification error. Two of the main factors include the effect of different limb positions and the presence of applied loads that may degrade the accuracy of movement intention detection.
In this work, using data from ten able-bodied subjects, we confirm that the variations in limb position associated with normal use can have a substantial impact on the robustness of EMG pattern recognition. In response to this, new classification methods based on the transient part of EMG data combined with the limb position information estimated with inertial measurement units (IMUs), are developed. The two-stage position-aware classification method, that uses an IMU-based automatic posture recognition followed by a position specific EMG-based movement classifiers, led to the lowest classification error. In particular, this approach reduced the average error respect to the solution using EMG only of 1.2% and 2.9% for the wrist and hand movement classification respectively.
File