Electroencephalogram (EEG) based brain–computer interface (BCI) is an augmented communication modality between the brain and computer that exclusively depends on noninvasively recorded neuro-electrical activity. To establish multiple control commands in BCI systems, it is essential to efficiently identify unique EEG activation patterns underlying multiple motor tasks. This article investigates the brain activation modulations elicited by imagined movement directions and the EEG features that characterize them. An experiment was conducted in which ten subjects imagined two-dimensional right-hand movement in the left, right, up, and down directions in the vertical plane. The frequency and phase characteristics of EEG rhythms are explored to identify the features that discriminate movement imagination toward multiple directions. The proposed decoding step consists of Fisher's analysis of instantaneous phase values to identify EEG channels with optimal discriminative information followed by extraction of phase-locking values and temporal-spatial features, which are then used for classification. The proposed approach offers significantly higher ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">p</i> < 0.05) performance than the state-of-the-art method while employing subject-specific time segments and channels. Hence, an average accuracy of 69.25% is obtained for binary classification of all pairs of movement directions with a performance increment of 8.33% compared to the conventional approaches using fixed channel-time paradigm. The results illustrate the potential use of noninvasive EEG signals in decoding imagined movement kinematics, and further research is needed to investigate and optimize the separability in multiple directions.