Özet:

Electromyographic (EMG) signals play an important role in controlling human-machine interactive smart hands prothes. The EMG signals that appear as a result of muscle activity contain specific information about the activity performed within themselves. Therefore, the enhancement of the functionality of smart hand proteins depends significantly on the proper analysis and interpretation of the EMG signal collected from the muscle area. In accordance with this concept, in the decision-making process of the smart hand protection movements, it is necessary to develop existing methods or suggest new methods that will prevail these methods in order to be reliably used the EMG signals. In the context of this study, with the aim of developing the analysis of multi-channel EMG signals, the method of excretion on the basis of multi-variable visual curve separation (CDGKA) has been presented as an alternative to traditional methods. The aim is to obtain more meaningful information from the EMG signal by using the CDGKA method that adapts the signal to the release modes. The performance of the CDGKA-based properties in distinguishing the different hands and fingers movements and in contrast to the different strength levels has been studied. For this purpose, the EMG signals of the low, medium and high strength levels of the ampute participants are now collected from the spots by applying the CDGKA method, the specific kip functions (ÖKF) have been obtained. It is classified using six different hands and fingers movements, the closest neighbor (k-NN), linear differentiation analysis (LDA) and support vector machine (SVM) classifiers. As a result of the classifications carried out by being trained and tested at the same level of strength (Scenario 1) and by being trained and tested at the same level of strength (Scenario 2) it was determined that the recommended CSDGKA-based properties, based on raw signal-based properties, are 10–15% for the 1 and up to 18% for the 2 scenarios.

Anahtar Kelimeler: