Hand Movement detection Using Empirical Mode Decomposition And Higher Order Spectra

With the industrialization of societies, the number of disabilities is increasing, and one of these disabilities can occur and severely affect the lives of individuals and society. There are several ways to control an artificial prosthesis, one of which is to use an electromyography signal. For this purpose, we use the dataset set available at the UCI database, which has 6 different hand movements in the form of free access. In this study, each signal is first decomposed into intrinsic mode, and each signal is converted to 8 IMF, and then, using high-order spectrum to show the changes. The results show that from the third IMF onwards, HOS patterns are repeated. The first and second IMFs are used as an input signal to artificial prosthesis and the feature of kurtosis, skewness and variance are extracted. The results have shown the accuracy of classification with the first IMF and SVM classifier is 98.26%.