EMG based classification of basic hand movements based on time-frequency features

Christos Sapsanis, George D. Georgoulas, Anthony Tzes

Abstract: This paper proposes an integrated approach for the identification of daily hand movements with a view to control prosthetic members. The raw EMG signal is decomposed into Intrinsic Mode Functions (IMFs) with the use of Empirical Mode Decomposition (EMD). A number of features are extracted in time and in frequency domain. Two different dimentionality methods are tested, namely the Principal Component Analysis (PCA) technique and the RELIEF feature selection algorithm. The outputs of the dimensionality reduction stage are then fed to a linear classifier to perform the detection task. The approach was tested on a group of young individuals and the results appear promising.