Brain-computer interface (BCI) is now being one of the interesting and attractive topics for researchers in recent years. Common spatial pattern is the most common approach applied in motor imagery (MI)-based BCI because of its effectiveness in differentiating between 2 different MI classes. The main objective of this article is to review using common spatial pattern combining with different methods such as Hilbert transform, error correction output coding, wavelet transform, joint approximate diagonalization, and others to extract features that can be used to differentiate between multiclass MI-based BCI. Data set 2a from BCI competition IV is used as an example for all authors. In addition, results of using various classifiers are demonstrated.