Perspective of signal processing-based on brain-computer interfaces using machine learning methods

Abstract

The application of artificial intelligence (AI) algorithms is an indispensable portion of developing brain-computer interfaces (BCI). With the continuous development of AI concepts and related technologies. AI algorithms such as neural networks play an increasingly powerful and extensive role in brain-computer interfaces. However, brain-computer interfaces are still facing many technical challenges. Due to the limitations of AI algorithms, brain-computer interfaces not only work with limited accuracy, but also can only be applied to certain simple scenarios. In order to explore the future directions and improvements of AI algorithms in the area of brain-computer interfaces, this paper will review and analyse the advanced applications of AI algorithms in the field of brain-computer interfaces in recent years and give possible future enhancements and development directions for the controversial parts of them. This review first presents the effects of different AI algorithms in BCI applications. A multi-objective classification method is compared with evolutionary algorithms in feature extraction of data. Then, a kind of supervised learning algorithm based on Event Related Potential (ERP) tags is presented to achieve a high accuracy in the process of pattern recognition. Finally, as an important experimental paradigm for BCI, a combined TFD-PSR-CSP feature extraction method, is explained for the problem of motor imagery. The 'Discussion' part comprehensively analyses the advantages and disadvantages of the above algorithms and proposes a deep learning-based artificial intelligence algorithm in order to solve the problems arising from the above algorithms.