A new subtle hand gestures recognition algorithm based on EMG and FSR

The paper proposes a new algorithm that combines the EMG (Electromyography) signals and pressure signals for detecting subtle hand gestures. First, the algorithm uses Myo armband to acquire the forearm EMG signals and the algorithm uses an array of five force sensitive resistors (FSRs) to acquire the muscle expansion pressure signals of the back of the hand. Then the algorithm has an activity detection on the collected signals. The algorithm has selected the top 50 best features of a gesture via the open source tool scikit-learn. After that the cross-validation is used to select the best parameter k of kNN(k-Nearest Neighbor) algorithm. Finally, the selected k of kNN algorithm is used for gesture classification in real time. The experiment shows that the algorithm presented in this paper has a classification accuracy of 96.05% for 21 pre-defined hand gestures, and the combination of EMG and FSR is significantly more accurate for subtle gesture recognition than the one using EMG only and using FSR only.