In the past few decades, brain-computer interfaces (BCIs) have emerged as new approaches for communication and control, bypassing the normal physiological neural pathways [1]. They have been treated as rehabilitation techniques, especially for severely disabled subjects, e.g., completely lockedin patients caused by amyotrophic lateral sclerosis (ALS), to replace or restore impaired communication or motor functions in clinical trials. A large number of studies have explored brain activities during movement execution where the kinematic parameters are correlated with the cortical processes [2][3]. More recently, decoding of brain signatures before motor execution, i.e., pre-movement states, has attracted more attentions since this information can be used to predict actions or anticipate the events. A BCI system can detect the user’s intention from non-invasive EEG signals and convert it into control signals to the external device. There have been closed-loop applications such as triggering peripheral electrical stimulation (ES) [4], controlling an orthosis [5], driving a wheelchair [6], and actuating a powered exoskeleton [7]. For these systems, an early detection of the movement intention from the user could be a prerequisite to close the perception-action loop. The cortical process of movement intention involves two modalities of EEG correlates, e.g., the motorrelated cortical potentials (MRCPs) and sensorymotor * Corresponding author: Weihai Chen (whchenbuaa@126.com) rhythms (SMRs) including event-related desynchronization/synchronization (ERD/S). Both of them have the advantage of high temporal precision and anticipation properties even for movement imagination or motor attempt, therefore compatible for the control signals to build an intuitive and natural BCI. SMR-based method has been used to predict human voluntary movement in real-time with a small false positive rate [8]. Another work by Pfurtscheller and SolisEscalante indicted that the beta rebound (ERS) over central middle line during foot motor imagery (MI) was relatively stable and reproducible in single EEG trials [9]. Although these SMR phenomenons are suitable to realize a a brain switch, long training is required for some users to learn to optimize the modulations of their brain patterns. On the other hand, movement-related cortical potentials (MRCP) has been employed in recent studies for the detection of pre-movement states, e.g., reaching movement [10], ankle dorsiflexion [11] and sitting/standing transition [12]. The MRCP is produced in corporation with movement planning and execution, with less training procedures. When generated in self-paced paradigms, it is termed as Bereitschaftspotential (BP) or readiness potential [13]. Being a slow cortical potentials (SCPs) close to DC, BP is elusive to present, which makes it challenging to detect from the background EEG activities. A possible way to boost the detection performance is to combine the MRCP-based and SMR-based methods. Interest has been devoted to studies about the upper-limb movement intention using both features [14], while the lower-limb work is still missing.