Abstarct: The problems that occur after stroke or diseases of the nervous system cause to speech or movement restricts in patients for a long time. Progress that have been made in the field of brain computer interface (BCI), make it possible to recognize and classify electrical or mextabolic activities of brain and convert them to control commands for a computer or a rehabilitation device. This thesis study focuses on the realization of an online cue baxsed Motor Imagery (MI) Brain Computer Interface (BCI) in order to control robotic arm. For this purpose, some signal processing and classification methods are investigated. Specifically, in feature extraction phase the wavelet packet is used to extract time and frequency features of input signals (related to two different motor imagery), DSLVQ and frequency search methods are applied as a procedure for feature selection, The performance of these methodologies is evaluated with the linear and nonlinear Support Vector Machines (SVM), Multilxayer Perceptron (MLP), Parzen, k-NN and Bayesian (NB) classifiers. The accuracy achieved between 67% to 90%, and the best kappa value is 0.8. At last, in order to improve the accuracy of classification, as an innovative procedure, classifier combination method is used for aggregate among the result of classifiers. By applying ordered weighted averaging (OWA) method and fuzzy integral baxsed methods (Sugeno and Choquet) the accuracy increased to 93%, 95% and 97.5% respectively. Also the required time for aggregation with OWA method is shorter than other combination methods. As the brain signals that are used related to two different imaginations so the result of classification is applied to control the motion of an elbow rehabilitation device. This rehabilitation device has been designed according to rehabilitation protocols of elbow in order to improve the speed of rehabilitation of patients. More over by applying a cable actuator, an appropriate simulation of elbow stiffness become possible. Among the advantages of this light and inexpensive mechanism is; generating smooth motion for joint with adjustable stiffness, compensation gravity effects, reduce the size of motors and in developed cases the ability of motivate number of lixnks with one power source simultaneously. More over back drivability of cable actuator generate a proper and safe interaction between user and device.