Systematic review of training environments with motor imagery brain–computer interface: Coherent taxonomy, open issues and recommendation pathway solution

The brain–computer interface (BCI) technique represents one of the furthermost active interdisciplinary study domains and includes a wide knowledge spectrum from a different disciplines such as medicine, neuroscience, machine learning and rehabilitation. The motor imagery (MI) technique based on BCI has been broadly applied in rehabilitation especially for upper limb motor movement where people with disabilities need to restore or improve their walking capability. Nowadays, virtual reality is a beneficial scheme for BCI users because it proposes a relatively cost-effective, safe way for BCI users to train and explain themselves in using BCI in a computer-generated environment earlier than in a real-life scenario. Depicting the whole picture for signal processing techniques and methods utilised in MI-based BCI training environments is difficult. In addition, numerous challenges and open issues regarding signal processing and pattern recognition exist in the literature of the current topic; however, to the best of our knowledge, this is the first attempt to highlight these challenges and open issues in signal processing methods, techniques and pattern recognition in smart BCI training environments. This work illustrates the effect of the theoretical perspectives associated with BCI works for research development in smart training environments. Consequently, this research copes with these issues via a systematic review protocol to help the large community of BCI users, especially people with disabilities. Fundamentally, four substantial databases, namely, IEEE, ScienceDirect, Scopus and PubMed contain a considerable amount of technical and scientific articles relevant to smart BCI training systems. A set of 375 articles is collected from 2010 to 2020 to reveal a clear picture and a better understanding of all the academic literature through a final set of 25 articles. In addition, this research provides the state of the art for signal processing, feature extraction, classification techniques and smart training environment characteristics for MI-based BCI applications. This study also reports the challenges and issues identified by the researchers as well as recommended solutions to solve the persistent problems. This study introduces the state-of-the art virtual and augmented reality environments as a smart platform and the neurofeedback schemes used for MI-based smart BCI training systems. Moreover, this study highlights for the first time 10 concepts of smart training in a virtual environment applied in MI and BCI, and investigates the evaluation of these concepts against the literature to gain only 45.55%. Collectively, the implication of this study will offer the opportunity of deploying an efficient smart BCI training system in terms of data acquisition and recording, pattern recognition and smart environment for BCI users and rehabilitation programmes.