Optimal Selection of Time Series Coefficients for Wrist Myoelectric Control Based on Intramuscular Recordings

Abstract

The Davies-Bouldin cluster validation technique has been utilized to determine the optimal time series parameters (model, order and number of coefficients) to be used in a myoelectric pattern recognition system based on intramuscular recordings. The data used in our longitudinal study were recorded intramuscularly (4 muscle channels) from an amputee subject over three different experimental days. During the recording sessions, the subject performed 4 wrist movements (pronation, supination, wrist flexion, wrist extension) in a way that was intuitive to her. The time series coefficients obtained from the EMG data (from AR, MA, and ARMA modeling) were used for generating feature spaces with 4 classes (1 per movement). The results showed that the optimal parameters didn't differ substantially (AR, order 4, 2 coefficients) for those obtained for other type of movements and surface EMG recordings. The use of the Davies-Bouldin technique during the analysis produced important information about the separability and consistency of the data across different days.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Oct 25, 2001
Accession Number
ADA411605

Entities

People

  • F. Sepulveda
  • M. C. Santa-cruz
  • R. Riso

Organizations

  • Aalborg University

Tags

DTIC Thesaurus Topics

  • Abstracts
  • Algorithms
  • Classification
  • Coefficients
  • Detection
  • Engineering
  • Feature Extraction
  • Frequency
  • Health Services
  • Military Research
  • Pattern Recognition
  • Prostheses And Implants
  • Recognition
  • Residual Limbs
  • Standards
  • Time Series Analysis
  • Validation

Readers

  • Brain and Cognitive Science; Experimental Psychology; Cognitive Neuroscience
  • Regression Analysis.
  • Robotics and Automation.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • Space