Çok Katmanlı Algılayıcı Yapay Sinir Ağı Kullanarak Packman Oyununda Yapılan El Hareketlerinin Sınıflandırılması

Electromyography is a noninvasive method that allows measurement of biological markers as a result of muscle activity. It is used as surface and needle electromyography according to the application purpose in two ways. In this study, seven different hand gestures (hand rest, hand punch, wrist bend, radial and ulnar deviation of the wrist) made by the individuals in the packman game in the „UCI Machine Learning Repository‟ database with open access over the internet were measured by using the data set of surface electromyogram signals tried to be classified. For this purpose, firstly, feature is extracted from data by discrete wavelet transform. Then, the extracted features were classified using the multi-layered sensor artificial neural network approach, which is widely used in the literature. In the classification process, artificial neural network was trained using simple cross validation algorithm, the algorithm and performance of the classifier were realized with Matlab2017a program.. The performance of the classifier has been investigated for the division of the data set at different rates and for different number of intermediate layers. The optimum network topology is obtained when the data set is divided by 20% -80% and the number of interlayers is 18, the highest performance is obtained as 91.67%.