In this study we present a modified version of t commercially-available myoelectric prosthesis (Myobock © , Ottobock) with the aim of providing a Brain-Machine Interface BMI-based sensorimotor control of this device. The new system uses as input the ElectroEncephaloGraphy (EEG) signals of the user as well as vibrations produced by a bracelet containing vibrating motors whose frequencies are proportional to the forces measured by Force-Sensitive Resistors (FSR) installed on the fingertips of the prosthesis. Four combinations of three different feature extraction methods (CSP, WD, GSO) have been used to construct the feature vectors of the EEG signals collected by two different recording systems with different number of electrodes during the experiments performed with seven able-bodied and four amputee subjects. The classification/prediction performances of three machine learning algorithms (Artificial Neural Network, Support Vector Machine with linear and Radial Basis Function kernels) were then tested. The reported results provide a proof of concept for the use of a wireless BMI to control the main types of movement of myoelectric prostheses using an EEG system with less electrodes rather than a research-grade system.