I. Introduction
One of the latest emerging technologies in the field of robotics and artificial intelligence is the system of brain- computer interfaces (BCI), also referred as Brain-Machine Interface (BMI), Mind-Machine Interface (MMI), or direct neural interface [1]. The BCI technology consists of the hardware that is able to recognize brain signals and the software that can effectively process these data [2]. As a result, successfully implemented BCI gives opportunity to control physical objects by decoding electrophysiological signals generated by human brain activities. Essentially, these kinds of signals are extracted from neurons in cortex when the electrodes are placed on a human scalp. These neural signals become useful as an input to measuring electronic devices due to the local field potential, which creates corresponding oscillatory wave [3]. After received brain signals are amplified, they are processed by analog to digital converter such that computer can perform further analysis. There are five commonly referred stages that neural signals are processed through: signal acquisition with preliminary noise reduction, signal preprocessing or enhancement, feature extraction, classification and output control interface [2]. Based on the communication channel between brain and the computer, this method of controlling devices without muscles and peripheral nerves can be implemented to solve problems with mobility impairment. For example, brain-actuated wheelchair is one of the most promising applications of BCIs that can help disabled people to move and interact with surrounded world through the intelligent robotics system [4]. In addition, various other invasive BCI systems have been developed to control external devices, e.g. computer cursors [5] and robotic prostheses/orthoses [6]. Moreover, in recent studies, BCIs have been used to control lower-body [7] and upper-body exoskeletons [8] for stroke and paraplegic recovery and rehabilitation via non-invasive approaches [9].