Developed Work and Current State (III)

Automatic detection on low-quality signals in HDEMG registers

HDEMG registers provide information from a greater proportion of the muscle of interest than conventional EMG. However, recording large numbers of channels often implies low-quality signals (or artifacts) caused by power line interference, short-circuits or variations in the baseline of the signals due to movement.

The presence of artifacts in the recorded databases suggested the implementation of a series of automatic detection methods which can also be used online in future HDEMG recordings. Three different methods have been designed and implemented:
 
  • An expert system based on binary decision trees whose thresholds were determined from a training database using Precision-Recall analysis.
  • A system based on the parameterization of hyperellipsoids obtained by an optimization process that employed evolutionary computation techniques.
  • A method based on the density of the data (the denser the area, the minor the probability of a sample of being an outlier).