Abstract
A robust system that classifies various hand gestures would greatly help those using prosthetic limbs. Recently, emphasis
has been placed on extracted features from the High Density - surface Electromyography (HD-sEMG) signals and
the size of segmentation windows which augment the recognition accuracy. This paper proposes a hand gestures
identification system, in which HD-sEMG signals are employed, and is supported by Force Myography (FMG) signals
for this mission. Several feature types have been extracted from FMG and HD-sEMG signals such as MEAN, RMS,
MAD, STD, and Variance, these features have been validated under some classifiers such as decision tree (DT), linear
discriminant analysis (LDA), support vector machine SVM, and k-nearest neighbor (KNN), in which results showing that
MEAN and RMS features are superior to others, while the best classifier is SVM. Several experiments have been achieved
by the MATLAB platform to validate the proposed system, in which, a database of HD-sEMG signals comprising 65
isometric hand gestures is employed, where two (8×8) electrodes and 9 force sensors are used to collect the FMG data.
This data was derived from 20 intact participants, the first preprocessing step was applied during the recording stage.
Ten gestures are chosen to be classified from the 65 hand gestures. Results show the success of the proposed system
while the classification accuracy arrived at 99.1%.