Towards Improving Motor Imagery Brain–Computer Interface Using Multimodal Speech Imagery

Abstract

Purpose: The brain–computer interface (BCI) based on motor imagery (MI) has attracted extensive interest due to its spontaneity and convenience. However, the traditional MI paradigm is limited by weak features in evoked EEG signal, which often leads to lower classification performance. Methods: In this paper, a novel paradigm is proposed to improve the BCI performance, by the speech imaginary combined with silent reading (SR) and writing imagery (WI), instead of imagining the body movements. In this multimodal (imaginary voices and movements) paradigm, the subjects silently read Chinese Pinyin (pronunciation) and imaginarily write the Chinese characters, according to a cue. Results: Eight subjects participated in binary classification tasks, by carrying out the traditional MI and the proposed paradigm in different experiments for comparison. 77.03% average classification accuracy was obtained by the new paradigm versus 68.96% by the traditional paradigm. Conclusion: The results of experiments show that the proposed paradigm evokes stronger features, which benefits the classification. This work opens a new view on evoking stronger EEG features by multimodal activities/stimuli using specific paradigms for BCI.