Hand gestures play a key role for controlling prosthetic and orthotic devices, enabling machine-human interaction, and telerehabilitation applications. Surface electromyography (sEMG) data from the hand and arm muscles have been used in numerous studies to identify hand movements. Several feature extraction techniques are used in the current work to classify EMG signals. Out of all the features, the autoregressive model performs the best. Techniques for feature selection are applied after feature extraction to minimize the dimensionality of the data which lowers computing costs and minimizes the chance of overfitting the data. This improves classification accuracy. Dimensionality reduction techniques include principal component analysis (PCA) and linear discriminant analysis (LDA). Additionally, various machine learning methods are also applied to the EMG signal classification process. We obtained 98.7% accuracy by using LDA as a feature selection method and support vector machine (SVM) as a classifier.