The brain-computer interface (BCI) technology enables direct interaction between the human brain and computers or other external devices by analyzing and decoding neural activity. It can be used as a means of information exchange or the restoration of motor functions and has been applied in communications, intelligent robot control, biomedicine, and neurorehabilitation, etc. Functional near-infrared spectroscopy (fNIRS), an optical imaging technique that can be used to detect changes in hemoglobin concentration within the cerebral cortex, has been employed recently in the development of noninvasive BCI. The development history, composition principles, key technologies, future development trends, limitations, and problems of fNIRS-BCI are reviewed systematically and in detail. Particularly, the feature-classification algorithm was analyzed comprehensively, and the result was compared with statistical data from its predecessors to summarize several valuable conclusions and opinions. This review is designed to provide a comprehensive and specific understanding of fNIRS-BCI and references and guidance.