Brain computer interface can realize the information interaction between human and outer devices and change “thinking” in the mind into real actions without the contact of the limb or neuromuscular system. Braincomputer interface is a novel product combining brain neuroscience and engineering technology, and also a new technique for clinical neural function rehabilitation and assistive motor control. It hopefully provides a brand new enhancing treatment and rehabilitation means for the patients losing the ability of language communication and body movement control partly or completely (such as stroke, spinal cord injury (SCI), amyotrophic lateral sclerosis (ALS) and etc.). However, current practical applications are still facing the technical bottlenecks of low information processing efficiency, long rehabilitation training time and poor control model generality. In this paper, these technical difficulties are reviewed. Then, taking the motor imagery BCI and BCI speller as examples, the possible model optimization strategies and solution schemes are introduced. The future development direction of BCI is discussed in the end of the paper.